Latest / 6 Ps in AI Pods (AI6P) / 🗣️ AISW #121: Nadina Dorene Lisbon, USA-based technical architect

Transcript

🗣️ AISW #121: Nadina Dorene Lisbon, USA-based technical architect · 6 Ps in AI Pods (AI6P)

• 0:00
  SPEAKER_00: Hi everyone, this is Karen Smiley.
• 0:02
  SPEAKER_00: Welcome to 6P's in AI Pods.
• 0:06
  SPEAKER_00: This post is part of our AI 6P interview series on AI, software and wetware.
• 0:13
  SPEAKER_00: Our guests share their experiences with using AI and how they feel
• 0:17
  SPEAKER_00: about AI using their data and content.
• 0:21
  SPEAKER_01: I'm delighted to welcome Nadina Lisbon from the USA as my guest today on AI,
• 0:26
  SPEAKER_01: Software and Wetware.
• 0:28
  SPEAKER_01: Nadina, thank you so much for joining me on this interview.
• 0:30
  SPEAKER_01: Please tell us about yourself, who you are and what you do.
• 0:34
  SPEAKER_02: Thank you so much for having me, Karen.
• 0:36
  SPEAKER_02: Yes, my name is Nadina Lisbon.
• 0:38
  SPEAKER_02: Currently, I work in software, specifically due to doing Salesforce at a partnership company.
• 0:45
  SPEAKER_02: And then from an and I'll see you next time.
• 0:46
  SPEAKER_02: Perspective,
• 0:47
  SPEAKER_02: I write TechSips with Nadina, which is a high -level weekly weekly blog on AI INTO.
• 0:53
  SPEAKER_02: Insights from either trending topics or reports that I've done research on.
• 0:58
  SPEAKER_01: Sounds good.
• 0:59
  SPEAKER_01: So tell us a little bit about what you're doing.
• 1:00
  SPEAKER_01: Bit about your level of experience with AI and machine learning and analytics.
• 1:05
  SPEAKER_01: How did you learn about AI and what is it that you focus on when you write about it
• 1:10
  SPEAKER_01: in Tech Tips with Nadina?
• 1:13
  SPEAKER_02: Yeah, I feel like when it comes to AI, I'm still a junior.
• 1:16
  SPEAKER_02: I'm still trying to get my feet wet from an AI perspective.
• 1:21
  SPEAKER_02: Since AI has come out, I have been heads and Heads Don't Really really doing two things.
• 1:26
  SPEAKER_02: At a previous role, I did a lot of AI strategy from a company person.
• 1:31
  SPEAKER_02: Perspective.
• 1:32
  SPEAKER_02: And so that's how I really got into to the AI side of it.
• 1:35
  SPEAKER_02: Side of things.
• 1:36
  SPEAKER_02: And what I focused on is more from a capability and for functionality perspective.
• 1:42
  SPEAKER_02: So really understanding what can AI do?
• 1:45
  SPEAKER_02: How is it different from automation?
• 1:47
  SPEAKER_02: I haven't gone much into the machine learning side of AI.
• 1:52
  SPEAKER_02: I've always had teams that work alongside me with that.
• 1:56
  SPEAKER_02: And then from an analytics approach,
• 1:58
  SPEAKER_02: it's really more,
• 1:59
  SPEAKER_02: I would say,
• 2:01
  SPEAKER_02: once the analytics are there, how can the AI read and interpret the data?
• 2:05
  SPEAKER_02: And is it actually the correct output.
• 2:08
  SPEAKER_02: Output that we're getting?
• 2:10
  SPEAKER_02: If we think of it from a tool perspective,
• 2:13
  SPEAKER_02: a lot of the AI is checked with your data type capabilities.
• 2:17
  SPEAKER_02: And then also I've been doing a lot of So, just understanding.
• 2:20
  SPEAKER_02: Understanding vibe coding and understanding how we can do
• 2:23
  SPEAKER_02: to essentially more more automation, but from an AI perspective to do with agents.
• 2:29
  SPEAKER_01: Can I ask, what's your educational background?
• 2:32
  SPEAKER_02: Yes, I have a master's in computational science and then my bachelor's is
• 2:37
  SPEAKER_02: in computer science in mathematics.
• 2:40
  SPEAKER_02: Originally,
• 2:41
  SPEAKER_02: when And I was was just leaving my master's,
• 2:45
  SPEAKER_02: I was trying to figure out, okay,
• 2:47
  SPEAKER_02: I have all this,
• 2:49
  SPEAKER_02: I like to say computer science education, what am I going to do with it?
• 2:54
  SPEAKER_02: I wasn't sure yet and then...
• 2:57
  SPEAKER_02: what I would be doing now if you'd asked me that question several years ago
• 3:01
  SPEAKER_02: that I would have gotten into Salesforce and doing that type of software.
• 3:06
  SPEAKER_02: It was more so that when I did my masters,
• 3:09
  SPEAKER_02: I did an internship at a, now you would call them, ISV company.
• 3:14
  SPEAKER_02: Basically companies that sell Salesforce as a service to other companies.
• 3:19
  SPEAKER_02: And that's where I learned about Salesforce.
• 3:21
  SPEAKER_02: I have since grown to So,
• 3:24
  SPEAKER_02: spread across the - the enterprise and learning about more technologies
• 3:28
  SPEAKER_02: and really understanding them from a capability perspective as it relates to CRM.
• 3:32
  SPEAKER_01: So with that educational background, you obviously are a well -positioned tech.
• 3:36
  SPEAKER_01: Technically to understand not just vibe coding, but real coding.
• 3:39
  SPEAKER_01: You probably don't.
• 3:40
  SPEAKER_01: Done your share of it and also to understand the math that is behind AI.
• 3:44
  SPEAKER_01: Yeah.
• 3:46
  SPEAKER_02: So it's fascinating because I actually started my career doing a lot of coding.
• 3:51
  SPEAKER_02: I wouldn't say I'm a specific hardcore Java or hardcore put whatever language in there.
• 3:59
  SPEAKER_02: I've done quite a few languages.
• 4:01
  SPEAKER_02: I've done Python.
• 4:02
  SPEAKER_02: I've done a in a little bit of a general Objective-C.
• 4:05
  SPEAKER_02: I feel like I'm a generalist when it comes to coding,
• 4:07
  SPEAKER_02: but the principles of object-oriented programming have really stuck with me.
• 4:12
  SPEAKER_02: I like to take that overarching principle and apply it to the different things I'm doing.
• 4:18
  SPEAKER_02: And so So the Vape Coding is...
• 4:20
  SPEAKER_02: is interesting because I feel like I'm I'm probably deleted some - adopter when it comes to it.
• 4:26
  SPEAKER_02: Not that I was nervous to start, but it was more so when I could...
• 4:32
  SPEAKER_02: code, I like to have a plan and I like to make things structured.
• 4:35
  SPEAKER_02: And so even now when I approach, let me work and see what the AI can do.
• 4:41
  SPEAKER_02: I start with that same mindset of What is the plan?
• 4:45
  SPEAKER_02: What are we trying to do?
• 4:46
  SPEAKER_02: What are we actually coding?
• 4:48
  SPEAKER_02: And so it's interesting working and I'm doing the the vibe coding with the AI,
• 4:52
  SPEAKER_02: because there's a lot of things where I wouldn't necessarily code it like
• 4:56
  SPEAKER_02: that today, but that's from my experience.
• 4:58
  SPEAKER_02: And so I realized I had to change how I was giving the prompts and the context to
• 5:03
  SPEAKER_02: the AI so that it can understand what I'm trying to do.
• 5:06
  SPEAKER_02: Trying to get to, because obviously you can't read my mind.
• 5:09
  SPEAKER_02: But at the same time,
• 5:11
  SPEAKER_02: I initially felt felt a bit slow because the time I took to explain it,
• 5:15
  SPEAKER_02: I probably could have written it faster.
• 5:18
  SPEAKER_01: Yeah, that's a good observation.
• 5:20
  SPEAKER_01: I hear that from a lot of people,
• 5:21
  SPEAKER_01: but you know,
• 5:22
  SPEAKER_01: understanding what what problem you want to solve and how you want to go about it
• 5:25
  SPEAKER_01: and what your constraints are, that's always been the hard part anyway.
• 5:29
  SPEAKER_02: Yeah, I think that's the age -old question.
• 5:31
  SPEAKER_02: Question, right?
• 5:32
  SPEAKER_02: What are we trying to do?
• 5:33
  SPEAKER_02: And then we get there and then there's always an iteration of the approach and then
• 5:37
  SPEAKER_02: and you're optimized.
• 5:38
  SPEAKER_02: Optimizing until there's no optimization left.
• 5:41
  SPEAKER_02: And then by the time you get to the end of it,
• 5:42
  SPEAKER_02: there's a new problem to solve, new code to develop.
• 5:46
  SPEAKER_01: Can you say a little bit about Salesforce?
• 5:48
  SPEAKER_01: You've mentioned this a couple of times and that's the area that you're working in.
• 5:50
  SPEAKER_01: And a lot of people may I think of Salesforce as well.
• 5:53
  SPEAKER_01: Salesforce as primarily a CRM and not really be aware of what its analytical capabilities are.
• 5:59
  SPEAKER_01: So how are you using and I'll see you next time.
• 6:02
  SPEAKER_01: Features within Salesforce as part of your job?
• 6:06
  SPEAKER_02: Yeah, I'll give a b*****g broad approach because I think for me,
• 6:09
  SPEAKER_02: Yeah, still sources a little - little different.
• 6:11
  SPEAKER_02: So let me give a little bit more background because people may not know.
• 6:15
  SPEAKER_02: When people look at Salesforce,
• 6:17
  SPEAKER_02: I think they look at it from a mainly the business lens of it's a CRM,
• 6:22
  SPEAKER_02: you get your contacts, your leads, your opportunities, you have your pipeline.
• 6:27
  SPEAKER_02: For me, I do Salesforce more from a technology approach and so indie salesforce.
• 6:33
  SPEAKER_02: Salesforce ecosystem,
• 6:34
  SPEAKER_02: there are different personas and then different slices of their products
• 6:38
  SPEAKER_02: and their clouds.
• 6:39
  SPEAKER_02: I will do my best to wrap all of it up immediately,
• 6:43
  SPEAKER_02: but you could literally talk about this for a day and a half and still not finish.
• 6:47
  SPEAKER_02: In the Salesforce world,
• 6:48
  SPEAKER_02: I'm a certified technical architect,
• 6:50
  SPEAKER_02: which is the highest level of certification you can get in Salesforce.
• 6:54
  SPEAKER_02: The number of Salesforce certified technical architects is less than 500 in the world.
• 7:00
  SPEAKER_02: But I still tell folks I'm a generalist because the products that I look at from
• 7:05
  SPEAKER_02: a Salesforce perspective,
• 7:07
  SPEAKER_02: it's sales, service, and really looking at the backend of Salesforce.
• 7:13
  SPEAKER_02: How do I connect it to other systems?
• 7:16
  SPEAKER_02: My experience is also a little more unique because I also was very much on
• 7:21
  SPEAKER_02: the development side of Salesforce.
• 7:23
  SPEAKER_02: So the coding, the integrations, working with their other companies like Mealsoft.
• 7:29
  SPEAKER_02: And from a salesforce perspective,
• 7:31
  SPEAKER_02: reports are in the standard product, but they also have marketing clothes.
• 7:36
  SPEAKER_02: Cloud and then they have commerce cloud.
• 7:39
  SPEAKER_02: And depending on what industry you're working on,
• 7:41
  SPEAKER_02: if you're doing government,
• 7:42
  SPEAKER_02: there's government cloud, which is heavily regulated to ensure data is protective.
• 7:47
  SPEAKER_02: If you're in financial services,
• 7:49
  SPEAKER_02: then you might be working from an insurance perspective or banking.
• 7:53
  SPEAKER_02: And so the slice and dice of the cloud, the persona.
• 7:58
  SPEAKER_02: And into work that that needs to be done can get technically deep,
• 8:04
  SPEAKER_02: but at the same time, it's very rewarding.
• 8:07
  SPEAKER_02: And so I think as you you're looking at Salesforce and folks look at it from
• 8:12
  SPEAKER_02: a business perspective and a technical perspective,
• 8:15
  SPEAKER_02: I've always been on the technical side.
• 8:17
  SPEAKER_02: So how can I get Salesforce to talk to other systems and how does Salesforce fit
• 8:21
  SPEAKER_02: within the overall enterprise?
• 8:23
  SPEAKER_02: But CRM perspective, that's Salesforce's bread and butter.
• 8:26
  SPEAKER_02: And most people know it from that perspective of contact management,
• 8:30
  SPEAKER_02: lead management, opportunity life cycle.
• 8:32
  SPEAKER_01: All right.
• 8:33
  SPEAKER_01: Thanks for explaining that.
• 8:34
  SPEAKER_01: I think that will be helpful as context for the rest of the conversation.
• 8:38
  SPEAKER_01: I'd like to hear a specific story some time that you had.
• 8:41
  SPEAKER_01: Have used an AI tool or a tool that included AI or machine learning features
• 8:46
  SPEAKER_01: that could be vibrated.
• 8:47
  SPEAKER_01: By coding.
• 8:48
  SPEAKER_01: It could be something related to Salesforce, whatever.
• 8:49
  SPEAKER_01: It could be something you've done on your own time.
• 8:51
  SPEAKER_01: Whatever it would be.
• 8:52
  SPEAKER_01: Be of interest to our audience.
• 8:55
  SPEAKER_01: I'd like to hear what worked well and what didn't work so well.
• 9:00
  SPEAKER_02: Yeah.
• 9:00
  SPEAKER_02: Okay.
• 9:00
  SPEAKER_02: So it's interesting.
• 9:02
  SPEAKER_02: Right now I've been using a lot of specifically Claude and Gemini.
• 9:09
  SPEAKER_02: So I'll give two use cases.
• 9:11
  SPEAKER_02: So, Gemini...
• 9:13
  SPEAKER_02: I use a lot in terms of of research perspective.
• 9:16
  SPEAKER_02: Perspectives before AI.
• 9:18
  SPEAKER_02: Anytime I'm researching something,
• 9:19
  SPEAKER_02: I need to go onto Google,
• 9:21
  SPEAKER_02: I would do the search,
• 9:22
  SPEAKER_02: I'd figure out the articles, then I need to go I'll read through a note.
• 9:26
  SPEAKER_02: Understand everything.
• 9:28
  SPEAKER_02: I use a lot now Gemini and specifically the Google Notebook.
• 9:32
  SPEAKER_02: Notebook and then Google has another problem.
• 9:34
  SPEAKER_02: Product called, I think it's Illuminate, where you can summarize all the data.
• 9:40
  SPEAKER_02: The one thing when I started using it,
• 9:43
  SPEAKER_02: I was very worried about,
• 9:44
  SPEAKER_02: okay, I've brought in all this data, I've summarized it and I'm reading it.
• 9:49
  SPEAKER_02: Is the output matching what I'm expecting?
• 9:52
  SPEAKER_02: And so a lot of,
• 9:55
  SPEAKER_02: I would say from my computer science principles,
• 9:58
  SPEAKER_02: I always follow garbage in garbage out.
• 10:00
  SPEAKER_02: And so I try to make sure I get the best resources,
• 10:02
  SPEAKER_02: even when I'm prompting the AI or asking it to do the research.
• 10:06
  SPEAKER_02: That it's pulling relevant data so that I can use it.
• 10:10
  SPEAKER_02: So that was one, eight percent.
• 10:10
  SPEAKER_02: A personal example.
• 10:12
  SPEAKER_02: Example from a work perspective.
• 10:16
  SPEAKER_02: I've been right now doing a lot of skill building with Claude.
• 10:22
  SPEAKER_02: So taking just typical tasks that I would normally do that would be something where,
• 10:28
  SPEAKER_02: you know, I need to write it out or create documentation.
• 10:30
  SPEAKER_02: And I've created it.
• 10:32
  SPEAKER_02: Created actual skills to either do the work,
• 10:36
  SPEAKER_02: analyze the work,
• 10:37
  SPEAKER_02: but there's still that human in the loop element of the skill
• 10:40
  SPEAKER_02: where I always have it paused to You ask me, okay.
• 10:44
  SPEAKER_02: Okay, review this.
• 10:45
  SPEAKER_02: And the reason why I do that is one,
• 10:49
  SPEAKER_02: I've seen things where here,
• 10:51
  SPEAKER_02: even though the - research was good, everything that I fed, the model was good.
• 10:55
  SPEAKER_02: We somehow got from point A to X with hallucinations.
• 11:00
  SPEAKER_02: And I couldn't really figure out why,
• 11:02
  SPEAKER_02: what had happened, like, how did you make these assumptions?
• 11:06
  SPEAKER_02: And as an artist.
• 11:07
  SPEAKER_02: Architect, I would say, trust but verify.
• 11:09
  SPEAKER_02: And so when I would and no reprograms.
• 11:11
  SPEAKER_02: Re-prompt to say,
• 11:13
  SPEAKER_02: help me understand how we got here,
• 11:15
  SPEAKER_02: it would have pulled in maybe a source that was not as credible,
• 11:18
  SPEAKER_02: even though I said, hey, just use these five sources.
• 11:22
  SPEAKER_02: Somehow it got a six or seven source pulled in as part of that data set.
• 11:26
  SPEAKER_02: And so, I always try and know.
• 11:28
  SPEAKER_02: Know and make sure that I have steps in place,
• 11:30
  SPEAKER_02: guardrails in place so that I can check before we move on to the next step so
• 11:35
  SPEAKER_02: that it's not one wasting time thinking about thoughts
• 11:40
  SPEAKER_02: that you shouldn't be thinking about.
• 11:41
  SPEAKER_02: And then to two, I like to understand, okay, how did you arrive to this conclusion?
• 11:47
  SPEAKER_02: Most of the time,
• 11:49
  SPEAKER_02: especially if I'm working with anything from a Salesforce perspective,
• 11:53
  SPEAKER_02: I would already know the answer.
• 11:54
  SPEAKER_02: So I usually do a reverse engineer to see if I can get it to the same level
• 12:00
  SPEAKER_02: of quality answer that I would expect to give for myself.
• 12:04
  SPEAKER_02: And then what would we do?
• 12:05
  SPEAKER_02: Do I need to do to get it to a level of confidence from an answer perspective that,
• 12:11
  SPEAKER_02: okay,
• 12:11
  SPEAKER_02: if somebody else on my team ran this skill,
• 12:15
  SPEAKER_02: they would be able to understand, yes, this is good.
• 12:18
  SPEAKER_01: Okay.
• 12:19
  SPEAKER_01: So these are skills that you're writing to use at work and then are you sharing them
• 12:22
  SPEAKER_01: with your team for them to use as well?
• 12:24
  SPEAKER_02: Yes.
• 12:24
  SPEAKER_02: So a lot of it is we try to create either frameworks for the skills or we have
• 12:30
  SPEAKER_02: a skill that is shared amongst us.
• 12:32
  SPEAKER_02: But with AI, sometimes the I will put your expenses on
• 12:36
  SPEAKER_02: Expecting, it's not always going to give you the same output.
• 12:38
  SPEAKER_02: And then depending on the input or what you want.
• 12:42
  SPEAKER_02: You've fed to it, it might be completely different.
• 12:44
  SPEAKER_02: So We're trying to essentially create a baseline of if we give it X, we expect Y.
• 12:51
  SPEAKER_02: And if we got Z, then we need to go back.
• 12:55
  SPEAKER_01: Okay.
• 12:55
  SPEAKER_01: So have you noticed that when you've developed any skills that they work now,
• 12:59
  SPEAKER_01: but then something in the model changes in a month later,
• 13:02
  SPEAKER_01: it doesn't work as well as it used to?
• 13:06
  SPEAKER_02: Yes, and I think it's it depends on what model I'm using.
• 13:10
  SPEAKER_02: It also depends on And again, the the underlying what I'm doing with that specific skill.
• 13:17
  SPEAKER_02: I have noticed that even with with Gemini, the Ma - model, I don't know when it should be.
• 13:23
  SPEAKER_02: Change happened.
• 13:23
  SPEAKER_02: Maybe it was like a month ago,
• 13:24
  SPEAKER_02: they changed a few things and the model itself has been not as,
• 13:29
  SPEAKER_02: I like to say, brilliant as it used to be.
• 13:31
  SPEAKER_02: And so I'm finding that I've had to kind of reset what I'm doing with it to ensure
• 13:37
  SPEAKER_02: that it's still at the level of quality I expect it to be.
• 13:41
  SPEAKER_01: You mentioned earlier that by the time you finished explaining
• 13:44
  SPEAKER_01: the AI tool what you want, you could have written it yourself.
• 13:48
  SPEAKER_01: Do you find that that's true also with the cloud skills?
• 13:51
  SPEAKER_01: If you think about the amount of time that you spend tweaking and monitoring them,
• 13:55
  SPEAKER_01: do you think it's less work than I'm doing the the actual researchers.
• 13:58
  SPEAKER_01: Research yourself?
• 13:59
  SPEAKER_02: Yeah,
• 14:00
  SPEAKER_02: I think upfront is definitely a cost because there needs to be time to actually,
• 14:06
  SPEAKER_02: again, for me,
• 14:07
  SPEAKER_02: I like to have I THOUGHT PROCESS process in mind
• 14:11
  SPEAKER_02: of I'm expecting this specific answer.
• 14:14
  SPEAKER_02: So I now need to, I think of it as how I would train a junior associate.
• 14:20
  SPEAKER_02: I can't expect them to get to my level of anxiety.
• 14:23
  SPEAKER_02: Answer on day one, because I don't think that's something you should expect of anyone.
• 14:28
  SPEAKER_02: It's like starting a new job.
• 14:30
  SPEAKER_02: They started on day one and you need them to be a rock star by day five.
• 14:33
  SPEAKER_02: It just doesn't happen.
• 14:34
  SPEAKER_02: And so I tried to to think about it that way.
• 14:38
  SPEAKER_02: Am I explaining to this LLM correctly?
• 14:41
  SPEAKER_02: Am I giving it the right context,
• 14:42
  SPEAKER_02: right prompts, right thinking so that they can get to that level?
• 14:46
  SPEAKER_02: And that does take some time.
• 14:48
  SPEAKER_02: So when I'm on a time crunch,
• 14:50
  SPEAKER_02: yeah, I'm going to do the work first because I have a hard time.
• 14:53
  SPEAKER_02: Deadline.
• 14:54
  SPEAKER_02: But once I get the response and I get some downtime, then I run it through again.
• 15:00
  SPEAKER_02: Because for me,
• 15:02
  SPEAKER_02: even when I go back to my experience of building code in the past,
• 15:06
  SPEAKER_02: I might have built something that took me a whole week.
• 15:09
  SPEAKER_02: If I rebuild it again,
• 15:11
  SPEAKER_02: it might take me three days and then I can figure out what the optimization is.
• 15:16
  SPEAKER_02: And so will I get that exact same project again?
• 15:19
  SPEAKER_02: No, but the principles of what I learned on how to build it,
• 15:23
  SPEAKER_02: then I would be able to apply that to a future project.
• 15:26
  SPEAKER_02: So I think of the same way with my thought skills, my Gemini gems and I think Chad G.
• 15:31
  SPEAKER_02: GPT, GPTs.
• 15:33
  SPEAKER_02: I'm giving in
• 15:34
  SPEAKER_02: it, D.
• 15:35
  SPEAKER_02: the rules, the instructions, it's going to change because things will naturally change.
• 15:42
  SPEAKER_02: But if I tweak it enough times and I do it enough times,
• 15:46
  SPEAKER_02: eventually it will get to a point where I can see it as stable and
• 15:49
  SPEAKER_02: then I'm just now building on what it needs to do.
• 15:52
  SPEAKER_02: I can also tell you tell immediately like maybe this model that I use was not
• 15:58
  SPEAKER_02: the best model or I needed to use a different way to give it the context clues
• 16:03
  SPEAKER_02: or whatever.
• 16:04
  SPEAKER_02: Whatever I'm working on.
• 16:06
  SPEAKER_01: So it sounds like you make a lot of good use of AI.
• 16:09
  SPEAKER_01: Do you use it for anything in your personal life, anything outside of work-related tasks?
• 16:15
  SPEAKER_02: Yeah, right? No.
• 16:16
  SPEAKER_02: Now, personally,
• 16:19
  SPEAKER_02: I've just been playing around with some vibe coding and then sometimes I'll use it
• 16:26
  SPEAKER_02: to do things like,
• 16:28
  SPEAKER_02: instead of searching on Google because it already has AI as well,
• 16:32
  SPEAKER_02: I also do more search.
• 16:34
  SPEAKER_02: Searches for or topics that I'm trying to...
• 16:37
  SPEAKER_02: trying to find deeper detail on or research on.
• 16:41
  SPEAKER_02: But even then,
• 16:42
  SPEAKER_02: I've noticed a difference between You know?
• 16:44
  SPEAKER_02: Searching Google with - without the AI search,
• 16:47
  SPEAKER_02: search in Google with the AI search,
• 16:49
  SPEAKER_02: and then searching using Gemini and saying do deep research.
• 16:55
  SPEAKER_02: Sometimes when I'm doing the tree searches,
• 16:59
  SPEAKER_02: this is when I have lots of time to really analyze it,
• 17:02
  SPEAKER_02: I've noticed that it I will not bring back back all the detail that it's searching
• 17:07
  SPEAKER_02: for.
• 17:07
  SPEAKER_02: Even though it's like a page one result, I would expect to see it there.
• 17:11
  SPEAKER_02: And then when I go to Google and do the AI search, same thing.
• 17:15
  SPEAKER_02: It kind of summarizes it, but it might not give all the page one and details to him.
• 17:20
  SPEAKER_02: I'm expecting.
• 17:20
  SPEAKER_02: And so and then I turn off.
• 17:22
  SPEAKER_02: Off and I said, no AI search.
• 17:24
  SPEAKER_02: Let me actually go visually look.
• 17:25
  SPEAKER_02: And it's like, why?
• 17:27
  SPEAKER_02: Then when I asked, why didn't you choose this specific topic?
• 17:31
  SPEAKER_02: Then it's more of it didn't think about it or it will include it next time.
• 17:36
  SPEAKER_02: And my goal is to figure out, okay, how can I make sure it includes it every time?
• 17:41
  SPEAKER_02: And so then I kind of go down a rabbit hole to figure out why.
• 17:44
  SPEAKER_02: Why was this not brought up or why?
• 17:46
  SPEAKER_02: This is page one of the Google search.
• 17:49
  SPEAKER_02: I would expect you would summarize it, but it didn't.
• 17:52
  SPEAKER_02: It did maybe the top three and I don't know So, if that's because of legal.
• 17:56
  SPEAKER_02: The SEO ranking or what is going on from the model's perspective
• 18:01
  SPEAKER_02: that is picking specific topics.
• 18:04
  SPEAKER_01: You mentioned that you have the Substack newsletter and that's actually how we met.
• 18:07
  SPEAKER_01: We met on Substack through the AI community there.
• 18:11
  SPEAKER_01: Do you ever use any AI tools as part of your writing process or for creating images
• 18:16
  SPEAKER_01: or anything like that?
• 18:18
  SPEAKER_02: Yes, I have used one called Napkin AI to do some of my images.
• 18:24
  SPEAKER_02: And then from a writing perspective I'm still very not nervous about it but even
• 18:30
  SPEAKER_02: when I do the research I still will write it by hand because I know
• 18:34
  SPEAKER_02: the data that's coming back is the data that is found
• 18:37
  SPEAKER_02: but again I don't know maybe it's the architect in me that I have to trust
• 18:41
  SPEAKER_02: and verify and so that's part of that process I just need to make sure that it's
• 18:45
  SPEAKER_02: in my own words because sometimes
• 18:48
  SPEAKER_02: when you like you'll have AI say hey can you summarize this
• 18:51
  SPEAKER_02: and it will give you back this long paragraph.
• 18:54
  SPEAKER_02: And I'm like, why are we having so much conversation?
• 18:57
  SPEAKER_02: They are overwhelming me with information.
• 18:59
  SPEAKER_02: And I will tell it,
• 19:00
  SPEAKER_02: can you give it to me,
• 19:01
  SPEAKER_02: concise, three bullet points or presenting this to an executive.
• 19:04
  SPEAKER_02: And it will for most of the time most part,
• 19:06
  SPEAKER_02: keep the context,
• 19:07
  SPEAKER_02: but then it will lose context within the middle of the trend for whatever reason.
• 19:11
  SPEAKER_02: Again, I know it's probably because of limits and windows and all those good things,
• 19:16
  SPEAKER_02: but I think that your voice itself, it's hard to train AI on it.
• 19:21
  SPEAKER_02: I know people, They'll feed it a breath.
• 19:23
  SPEAKER_02: Brand and it'll get pretty close, but there's just certain human things that you should do.
• 19:28
  SPEAKER_02: And even if it was will help me with the research, I still try to write it by hand.
• 19:34
  SPEAKER_02: I have my own template that I use.
• 19:35
  SPEAKER_02: There's not a lot thinking for the template part,
• 19:37
  SPEAKER_02: which I did have AI help me figure out what's a concise template that's easy
• 19:41
  SPEAKER_02: to read,
• 19:41
  SPEAKER_02: but in actual content,
• 19:43
  SPEAKER_02: I try to write by hand so that it's my voice and again, again.
• 19:48
  SPEAKER_02: I don't know.
• 19:49
  SPEAKER_02: I don't know if I would ever be able to train the AI to a hundred percent right like me.
• 19:52
  SPEAKER_02: Me because even and my voice will will evolve over time as I learn new things
• 19:57
  SPEAKER_02: in different ways.
• 19:58
  SPEAKER_02: And so Well, that's...
• 20:00
  SPEAKER_02: a conversation in itself about how well AI could write as a person or as me.
• 20:05
  SPEAKER_02: But yeah,
• 20:06
  SPEAKER_02: I have used some AI to do some of my images and then I'll use Google images as well
• 20:11
  SPEAKER_02: for some of those topics.
• 20:13
  SPEAKER_01: It sounds like you're making a lot of good use of AI tools for a lot of different things.
• 20:17
  SPEAKER_01: Are there any things that you specifically avoid using it for other than your writing?
• 20:22
  SPEAKER_01: Would you to avoid grief.
• 20:24
  SPEAKER_01: Creating music or videos or writing a personal story.
• 20:28
  SPEAKER_01: Personal email or something like that?
• 20:30
  SPEAKER_02: Yeah, I think it's interesting because I think anything where it's somebody's leg.
• 20:36
  SPEAKER_02: Livelihood always This makes me me a bit wary so a musician
• 20:40
  SPEAKER_02: for example has spent years and years learning to trade learning music
• 20:44
  SPEAKER_02: and AI can yeah I can create something in minutes but as a person I have just
• 20:51
  SPEAKER_02: a little bit enough to understand music but I'm not at that level of mastery
• 20:56
  SPEAKER_02: where I could listen and see you later.
• 20:58
  SPEAKER_02: Like oh yeah this is the greatest thing and I should go publish this today it's more
• 21:02
  SPEAKER_02: of this was a fun project I don't see AI becoming this thing where it's going
• 21:08
  SPEAKER_02: to help people, yes,
• 21:09
  SPEAKER_02: but at the same time,
• 21:10
  SPEAKER_02: there's just certain things that people do a lot of people really well.
• 21:13
  SPEAKER_02: Well.
• 21:14
  SPEAKER_02: They've taken the time to master it.
• 21:16
  SPEAKER_02: And so having AI try to replicate that mastery,
• 21:21
  SPEAKER_02: it feels a bit weird to be like,
• 21:23
  SPEAKER_02: oh yeah, today I'm publishing music because I God Belied His Ingredients.
• 21:26
  SPEAKER_02: And greatest model that can and synthesize all these.
• 21:29
  SPEAKER_02: These amazing people and do it.
• 21:30
  SPEAKER_02: It just feels like something you shouldn't do.
• 21:32
  SPEAKER_02: It's similar to artists when you're creating paintings, right?
• 21:37
  SPEAKER_02: Somebody spent time learning how to paint, learning brushstrokes, learning.
• 21:42
  SPEAKER_02: They look at a blank canvas and they can see what that image looks like.
• 21:46
  SPEAKER_02: And I understand that to a sense because I can look at Salesforce and see,
• 21:51
  SPEAKER_02: oh yeah, this is a problem.
• 21:52
  SPEAKER_02: I can tell you a hundred percent how to solve it.
• 21:54
  SPEAKER_02: I don't think about it.
• 21:55
  SPEAKER_02: It's like breeding for me.
• 21:56
  SPEAKER_02: And so to So then try and have a look at it.
• 21:58
  SPEAKER_02: An AI replicate that is just for lack of better words, weird.
• 22:03
  SPEAKER_01: Okay.
• 22:03
  SPEAKER_01: Thanks for sharing that.
• 22:04
  SPEAKER_01: So one concern that has come up is where the AI and machine learning tools get
• 22:12
  SPEAKER_01: the data that they use for we're creating these models.
• 22:15
  SPEAKER_01: Models that can and generate.
• 22:17
  SPEAKER_01: Generate and new combinations.
• 22:18
  SPEAKER_01: Combinations, new texts, new works.
• 22:21
  SPEAKER_01: And a lot of times they'll use information information to people.
• 22:23
  SPEAKER_01: People put online.
• 22:24
  SPEAKER_01: For instance,
• 22:25
  SPEAKER_01: on Substack,
• 22:26
  SPEAKER_01: we have an option to disable allowing the AI bots to scrape our newsletters.
• 22:32
  SPEAKER_01: And I'm curious if you've We've got that setting to turned on or turned off.
• 22:36
  SPEAKER_01: Do you allow bots to scrape your newsletter?
• 22:39
  SPEAKER_02: Yeah, so I learned about it a little too late.
• 22:42
  SPEAKER_02: So I think they've already been scraping.
• 22:44
  SPEAKER_02: So I didn't turn mine off.
• 22:46
  SPEAKER_02: At the same time,
• 22:48
  SPEAKER_02: it's interesting because I think AI is definitely running out
• 22:52
  SPEAKER_02: of what can be fed now,
• 22:54
  SPEAKER_02: what information instead of getting this information from all over.
• 22:59
  SPEAKER_02: It's what one of the reasons why I'm also trying to move off of the Gmail ecosystem.
• 23:03
  SPEAKER_02: It's the same thing.
• 23:04
  SPEAKER_02: One day,
• 23:05
  SPEAKER_02: everything was just like,
• 23:07
  SPEAKER_02: yeah, AI can now access your Gmail, your drive and everything you have.
• 23:13
  SPEAKER_02: And so I did turn it off,
• 23:14
  SPEAKER_02: but again, I felt like I was too late because AI, once it's on, it does it quickly.
• 23:19
  SPEAKER_02: And I have a lot of my personal writing in there.
• 23:21
  SPEAKER_02: There's a mini panic.
• 23:22
  SPEAKER_02: I started moving everything and then I was like,
• 23:24
  SPEAKER_02: okay, I need to go to a different email provider, which I'm slowly moving towards.
• 23:29
  SPEAKER_02: But the fundamental question I was thinking about is how do I really protect
• 23:37
  SPEAKER_02: at least myself digitally online?
• 23:40
  SPEAKER_02: Because now throughout the years,
• 23:43
  SPEAKER_02: if you Google my name, there's things where I've spoken at different conferences.
• 23:48
  SPEAKER_02: Things that I've written.
• 23:49
  SPEAKER_02: and I'm sure with Google,
• 23:50
  SPEAKER_02: good AI,
• 23:51
  SPEAKER_02: you probably could spoof me at this point
• 23:54
  SPEAKER_02: because I've already put out too much content on myself.
• 23:58
  SPEAKER_02: And so fundamentally there needs some type of guardrails to say what's in
• 24:03
  SPEAKER_02: and what's out,
• 24:04
  SPEAKER_02: but even just a toggle,
• 24:06
  SPEAKER_02: how do we know, how do we trust that the bots won't try to still take it?
• 24:11
  SPEAKER_02: Yes, it's off from an API perspective,
• 24:14
  SPEAKER_02: but who's the person that's making sure that it's actually off, right?
• 24:18
  SPEAKER_02: It's a toggle on the UI, but is the API actually LOCKDOWN.
• 24:22
  SPEAKER_02: Down that if the agent or whoever STRAPES IT DUDE.
• 24:26
  SPEAKER_02: They actually get an error message?
• 24:28
  SPEAKER_02: I think that's fundamentally where we are from the security perspective right now.
• 24:32
  SPEAKER_02: I'm going to continue to try and not have my things scraped,
• 24:36
  SPEAKER_02: but at the same time,
• 24:37
  SPEAKER_02: there's only so much I can do and there's a little bit of of,
• 24:42
  SPEAKER_02: by the time I f*****g found out about the Substack one,
• 24:44
  SPEAKER_02: I think it probably had all my posts at that point.
• 24:47
  SPEAKER_02: So I didn't turn it off because at that point it looked like it will get what, maybe one?
• 24:53
  SPEAKER_02: It felt like I was too late.
• 24:55
  SPEAKER_02: Kind of similar with Gmail.
• 24:57
  SPEAKER_02: By the time I had turned it off,
• 24:59
  SPEAKER_02: it had already probably gotten a good understanding of me from years of history,
• 25:04
  SPEAKER_02: years of whatever I have in my drive.
• 25:07
  SPEAKER_02: And so I think fundamentally Yeah,
• 25:09
  SPEAKER_02: definitely where - possible, yes,
• 25:11
  SPEAKER_02: you should prevent it from taking your information
• 25:15
  SPEAKER_02: because I don't think AI is out here when it's feeding information to people.
• 25:19
  SPEAKER_02: Like if you searched me or search my work that it's like,
• 25:23
  SPEAKER_02: yeah, credit Nadina for this work that I have now taken and summarized for you.
• 25:29
  SPEAKER_02: I think if we could get to that,
• 25:30
  SPEAKER_02: maybe people would be at ease, but then at the same time, AI is very good acts.
• 25:36
  SPEAKER_02: Taking a bunch of information,
• 25:38
  SPEAKER_02: putting it together and making it look like it's possibly something it came up with,
• 25:44
  SPEAKER_02: but nothing's new.
• 25:46
  SPEAKER_02: Everything's taken from somewhere.
• 25:48
  SPEAKER_02: It's gotten it from somewhere.
• 25:49
  SPEAKER_02: So it's kind of interesting.
• 25:51
  SPEAKER_02: But yes, long story short, I didn't turn off my Substack one.
• 25:54
  SPEAKER_02: I turned off my Gmail one.
• 25:57
  SPEAKER_02: I hope you can see.
• 25:59
  SPEAKER_02: That in time we have better security around APIs
• 26:02
  SPEAKER_02: because APIs fundamentally have been very open and a lot of
• 26:06
  SPEAKER_02: the protections have been from a UI perspective,
• 26:09
  SPEAKER_02: but from an AI perspective, once it has access via What's your...
• 26:15
  SPEAKER_02: AI or true a MCP MCP, a model context protocol, the skies is the limits.
• 26:22
  SPEAKER_01: Yeah.
• 26:22
  SPEAKER_01: And that's exactly one of the reasons Why?' Why being opted in by default
• 26:27
  SPEAKER_01: when they turn on these new settings is a problem because the minute that it's on,
• 26:32
  SPEAKER_01: even if you turn it off within an hour,
• 26:35
  SPEAKER_01: it's still potentially already been scraped up and folded in.
• 26:38
  SPEAKER_01: And once it gets trained into an AI model, it doesn't come back out, you know?
• 26:44
  SPEAKER_02: Yeah,
• 26:44
  SPEAKER_02: and I got the default opt-in is where I've always had a problem regardless from
• 26:48
  SPEAKER_02: an AI perspective, just in general.
• 26:50
  SPEAKER_02: I don't like the default opt-in.
• 26:52
  SPEAKER_02: I think people should be able to opt into what they need.
• 26:56
  SPEAKER_02: So it should start us off and then you should be able to turn it on
• 26:59
  SPEAKER_02: because Not everyone's getting realized that it's happened.
• 27:04
  SPEAKER_02: And then by the time it has happened, it's too late.
• 27:06
  SPEAKER_02: And then you're in this crisis of, okay,
• 27:08
  SPEAKER_02: I can turn it off, but it's probably already taken everything it wanted to take.
• 27:12
  SPEAKER_02: And so it's only net new now.
• 27:14
  SPEAKER_01: It was about two years ago now where LinkedIn did something like this,
• 27:18
  SPEAKER_01: where they were opting all of us in by default to everything we'd ever put together.
• 27:22
  SPEAKER_01: Put into the system up to that point.
• 27:24
  SPEAKER_01: And they said, But you have an offer.
• 27:26
  SPEAKER_01: Option where you can now opt out from us taking any of your new stuff.
• 27:29
  SPEAKER_01: Do you remember when that happened?
• 27:31
  SPEAKER_01: Did that affect you?
• 27:32
  SPEAKER_01: Did it come to your attention?
• 27:32
  SPEAKER_01: Because they were not exactly upfront about it.
• 27:35
  SPEAKER_02: Yeah, it came up.
• 27:37
  SPEAKER_02: It was strange how I found out about it because I found out on LinkedIn through
• 27:41
  SPEAKER_02: a group that was like, make sure you opt out.
• 27:44
  SPEAKER_02: And I was like, opt out of what?
• 27:46
  SPEAKER_02: And when I phoned, found out, it's already taken the old stuff, so I can't opt out of the old.
• 27:51
  SPEAKER_02: I thought I was not reading it correctly.
• 27:53
  SPEAKER_02: You're only opting out of the net view.
• 27:56
  SPEAKER_02: Which I'm going to use the word interesting approach because again,
• 28:00
  SPEAKER_02: and one would have used expected to be able to opt out of everything,
• 28:06
  SPEAKER_02: but it was more of actually, no, you don't have that choice.
• 28:09
  SPEAKER_02: You get to opt out of net new.
• 28:11
  SPEAKER_02: Just assume this is all ors.
• 28:13
  SPEAKER_02: And I think that's one of the things with social media too,
• 28:16
  SPEAKER_02: where you join something like Facebook,
• 28:20
  SPEAKER_02: Instagram, and if you're not paying for these p*****s.
• 28:23
  SPEAKER_02: Products.
• 28:24
  SPEAKER_02: No, you are the product.
• 28:25
  SPEAKER_02: And what you've posted on there,
• 28:30
  SPEAKER_02: I'm sure it's in some terms and conditions that nobody's read
• 28:33
  SPEAKER_02: that it can either read it or is it it's become part of now their training set
• 28:38
  SPEAKER_02: for machine learning or something else.
• 28:40
  SPEAKER_02: And it's not readily available where somebody would be able to make that decision
• 28:46
  SPEAKER_02: because they wouldn't even know to think,
• 28:49
  SPEAKER_02: oh, I should really check and see what I'm that I'm logging into right now.
• 28:53
  SPEAKER_02: Right?
• 28:53
  SPEAKER_02: Most people are like, oh yeah, new social media, let me get on first.
• 28:56
  SPEAKER_02: Oh yeah.
• 28:57
  SPEAKER_02: Click, click, click I've
• 28:58
  SPEAKER_02: already passed through the terms and conditions.
• 29:00
  SPEAKER_02: And so, yeah, it's just an interesting question in the whole opt-in, opt-out conversation.
• 29:08
  SPEAKER_01: So you mentioned that you I use Gemini.
• 29:10
  SPEAKER_01: Gemini and Quad.
• 29:12
  SPEAKER_01: I'm curious if you feel like Google and Anthropic have been transparent with you as
• 29:18
  SPEAKER_01: a user about where they got the data that was used for training those tools.
• 29:25
  SPEAKER_02: I think transparency is an interesting word.
• 29:28
  SPEAKER_02: Oh well I see a lot more of, hey,
• 29:30
  SPEAKER_02: we're updating our terms and conditions,
• 29:32
  SPEAKER_02: but I know just because I've been following,
• 29:35
  SPEAKER_02: Anthropic had the lawsuit for the books that it had stolen.
• 29:41
  SPEAKER_02: Google.
• 29:42
  SPEAKER_02: Was trained on Reddit's data and we don't even know what is it?
• 29:47
  SPEAKER_02: Is in there because they're there was a whole mess of things that was happening
• 29:51
  SPEAKER_02: from Google perspective.
• 29:53
  SPEAKER_02: But Google itself is a search engine.
• 29:55
  SPEAKER_02: I just think about it and how much data it's now had.
• 29:59
  SPEAKER_02: And when it had that whole, yeah,
• 30:02
  SPEAKER_02: you're automatically opt-in surprise where it could read your Gmail,
• 30:06
  SPEAKER_02: your drive, Gemini is just running through everything.
• 30:09
  SPEAKER_02: A lot of ideas.
• 30:10
  SPEAKER_02: Data that it got, I don't know if it got it fairly.
• 30:14
  SPEAKER_02: And I don't know where to go.
• 30:15
  SPEAKER_02: As these models are getting data,
• 30:16
  SPEAKER_02: what is considered fair in terms of there's some line
• 30:21
  SPEAKER_02: that people are probably coming up against, probably have crossed it.
• 30:25
  SPEAKER_02: I think initially when it first came out,
• 30:28
  SPEAKER_02: especially with OpenAI, people didn't even ask the question, right?
• 30:31
  SPEAKER_02: It's like, oh, it has all this data.
• 30:33
  SPEAKER_02: It's huge.
• 30:34
  SPEAKER_02: It's going to do everything.
• 30:35
  SPEAKER_02: And then it's going to give you this exact answer.
• 30:38
  SPEAKER_02: But where did the data come from?
• 30:39
  SPEAKER_02: I don't think people started to ask that question until it was,
• 30:42
  SPEAKER_02: I feel like it was a little too late, but now So, at least the question is.
• 30:47
  SPEAKER_02: Is being asked.
• 30:47
  SPEAKER_02: So in the beginning to know no,
• 30:49
  SPEAKER_02: I don't think they were upfront on how they got the data
• 30:53
  SPEAKER_02: and how they were training these models.
• 30:56
  SPEAKER_02: At a previous company, actually, when we were looking at doing, or AIBC.
• 31:00
  SPEAKER_02: We specifically didn't do any of the public models.
• 31:05
  SPEAKER_02: So we had a private modeling.
• 31:07
  SPEAKER_02: Model internally that you could access and you could not access any of the GPT, the Gemini.
• 31:13
  SPEAKER_02: You couldn't access any of it from your work computer.
• 31:16
  SPEAKER_02: So it was just blocked.
• 31:17
  SPEAKER_02: That was quite an interesting time for people
• 31:19
  SPEAKER_02: because they felt like they weren't able to So, do what he needed to do.
• 31:23
  SPEAKER_02: Do.
• 31:23
  SPEAKER_02: But at the same time,
• 31:24
  SPEAKER_02: there were so many leaks of companies putting data in public models
• 31:28
  SPEAKER_02: and people not understanding that once you put it in there, it's out there.
• 31:32
  SPEAKER_02: And I think now people are a bit more secure.
• 31:36
  SPEAKER_02: They have their own models internally.
• 31:38
  SPEAKER_02: There's more tooling around.
• 31:40
  SPEAKER_02: I I have my work clock.
• 31:42
  SPEAKER_02: Cloud.
• 31:42
  SPEAKER_02: I have my personal cloud and those two should not meet.
• 31:45
  SPEAKER_02: The work should go into work.
• 31:46
  SPEAKER_02: The personal should go into personal and do not mix them.
• 31:49
  SPEAKER_02: I think there's more understanding there.
• 31:51
  SPEAKER_02: Same with Google, same with ChatGPT.
• 31:54
  SPEAKER_02: But I think initially when everything came out,
• 31:56
  SPEAKER_02: the excitement of,
• 31:58
  SPEAKER_02: I have a tool that can help me write an email,
• 32:01
  SPEAKER_02: which which is what a lot of people think.
• 32:03
  SPEAKER_02: People were using for,
• 32:04
  SPEAKER_02: or I know can have have this short succinct email for corporate America,
• 32:09
  SPEAKER_02: or it can summarize these meeting transcripts for me and then I can get
• 32:13
  SPEAKER_02: the action items.
• 32:15
  SPEAKER_02: I don't think people were worrying too much
• 32:16
  SPEAKER_02: about how these models were being trained,
• 32:18
  SPEAKER_02: but at the same time, It was always the top of my head.
• 32:21
  SPEAKER_02: Top of mind for me.
• 32:23
  SPEAKER_02: Like, how are you getting this data?
• 32:24
  SPEAKER_02: How do you know so much?
• 32:26
  SPEAKER_01: So it's one thing when we have a choice about using it.
• 32:29
  SPEAKER_01: I'm going to need it for tool at work or at home.
• 32:33
  SPEAKER_01: In a lot of cases,
• 32:34
  SPEAKER_01: our data is being used just because we are consumers,
• 32:38
  SPEAKER_01: we are members of the public,
• 32:40
  SPEAKER_01: and we interact with government and commercial systems.
• 32:43
  SPEAKER_01: Do you know of any cases where your personal data has been and used by by one of these systems?
• 32:49
  SPEAKER_01: Obviously without disclosing any specific personal data.
• 32:52
  SPEAKER_02: I don't have a specific case of my DataBeans I'll say taken and use without.
• 32:59
  SPEAKER_02: Without my intent,
• 33:01
  SPEAKER_02: but doesn't mean that it doesn't exist because like I said,
• 33:04
  SPEAKER_02: there's so many tools out there now.
• 33:07
  SPEAKER_02: Even They're doing something.
• 33:09
  SPEAKER_02: Something like you're trying to research someone,
• 33:12
  SPEAKER_02: there's so much information that can come back.
• 33:14
  SPEAKER_02: It was interesting because I did do a reverse,
• 33:17
  SPEAKER_02: do you know who I am, search my name and what comes back.
• 33:20
  SPEAKER_02: And it was fascinating, the information that came back.
• 33:24
  SPEAKER_02: A lot of it, again,
• 33:24
  SPEAKER_02: was public information,
• 33:26
  SPEAKER_02: information that if you do a Google search, you could find it on me.
• 33:29
  SPEAKER_02: But some of it was, again,
• 33:30
  SPEAKER_02: page three,
• 33:31
  SPEAKER_02: page four, things that you really have to go looking to find that information.
• 33:35
  SPEAKER_02: So it was interesting how that that input it already.
• 33:39
  SPEAKER_02: Already ended up in the model.
• 33:40
  SPEAKER_02: Again, it was public.
• 33:41
  SPEAKER_02: I probably put it out there.
• 33:43
  SPEAKER_02: It's kind of similar to Facebook.
• 33:44
  SPEAKER_02: Like when I first started, I didn't know I was that's not supposed to happen.
• 33:48
  SPEAKER_02: Have my Facebook public.
• 33:49
  SPEAKER_02: I quickly changed that, but again, the data was already out there.
• 33:52
  SPEAKER_02: So I'm sure I and got some of that info.
• 33:54
  SPEAKER_02: Info as well.
• 33:55
  SPEAKER_02: But yeah,
• 33:56
  SPEAKER_02: I don't have a specific case where my information was taken,
• 33:59
  SPEAKER_02: but I'm sure that it's probably out there and it's probably a clone
• 34:04
  SPEAKER_02: of myself somewhere running around right now.
• 34:06
  SPEAKER_01: So, all right.
• 34:07
  SPEAKER_01: Right.
• 34:08
  SPEAKER_01: Do you know any companies?
• 34:09
  SPEAKER_01: Company that you gave your data to that actually alerted you
• 34:12
  SPEAKER_01: that they would be using it for AI A .I., or Machine Learning?
• 34:15
  SPEAKER_01: Learning?
• 34:16
  SPEAKER_02: It was interesting.
• 34:17
  SPEAKER_02: When I was I was interviewing last week.
• 34:19
  SPEAKER_02: Last year,
• 34:20
  SPEAKER_02: there were a lot of companies where there were like,
• 34:23
  SPEAKER_02: for those first interviews, it was more like they were asking and you could opt out.
• 34:29
  SPEAKER_02: But again,
• 34:29
  SPEAKER_02: it was already an opt -in,
• 34:31
  SPEAKER_02: so you were already checked yes and you had to opt out specifically,
• 34:34
  SPEAKER_02: but it was Hey,
• 34:35
  SPEAKER_02: for AI training purposes, we may record this interview to use to train or our model.
• 34:41
  SPEAKER_02: And so I really started paying attention when I was looking at what are they asking
• 34:47
  SPEAKER_02: of me in these interviews and why are we recording it and why are we using it
• 34:52
  SPEAKER_02: to train AI?
• 34:54
  SPEAKER_02: I've had friends actually where the first this interview was nice.
• 34:57
  SPEAKER_02: Not with a human, it was with an AI.
• 35:00
  SPEAKER_02: And so I joke and I said, you should send your AI to do the interview.
• 35:04
  SPEAKER_02: If they're going to send an AI to do the interview, kind of black mirror.
• 35:08
  SPEAKER_02: It's interesting that the first interview would be AI.
• 35:11
  SPEAKER_02: A lot of the applications were already being sorted by AI.
• 35:15
  SPEAKER_02: When were you going to see
• 35:16
  SPEAKER_02: See a human in that entire process?
• 35:18
  SPEAKER_02: So it's interesting now because I have friends who are interviewing and they say,
• 35:22
  SPEAKER_02: yes, some of the roles, the first call is not with a person.
• 35:25
  SPEAKER_02: It's with an AI that
• 35:26
  SPEAKER_02: That will ask you questions.
• 35:27
  SPEAKER_02: I didn't participate in any of those, nor am I planning to participate in any of them.
• 35:32
  SPEAKER_02: I feel like if we're going to interview for a job, I'm not asking an AI for a role.
• 35:38
  SPEAKER_02: From a company perspective, I understand that they're trying to to optimize the
• 35:43
  SPEAKER_02: Recruiting process.
• 35:44
  SPEAKER_02: We can't talk to so many candidates at a time,
• 35:46
  SPEAKER_02: but at the same time,
• 35:47
  SPEAKER_02: there's just certain things a human touch in the first part of an interview kind
• 35:51
  SPEAKER_02: of tells a lot or not a lot about it.
• 35:54
  SPEAKER_02: A company.
• 35:55
  SPEAKER_02: Like if you're going to use AI in the beginning, is AI good.
• 35:59
  SPEAKER_02: Going to be the thing that I'm going to be doing performance reviews on?
• 36:02
  SPEAKER_02: Is AI going to be the thing that I'm talking to all the time?
• 36:04
  SPEAKER_02: When am I going to get a human to talk to and have an actual conversation?
• 36:10
  SPEAKER_02: So That was kind of interesting when I was interviewing
• 36:12
  SPEAKER_02: and even now my friends are interviewing that the first specific call is with AI.
• 36:18
  SPEAKER_01: And using AI in hiring and performance evaluation is It's problematic in a lot of ways.
• 36:24
  SPEAKER_01: Lot of ways because there's a lot of bias built into the old data
• 36:26
  SPEAKER_01: that they would have used for training those systems.
• 36:30
  SPEAKER_01: So it's a little disconcerting.
• 36:32
  SPEAKER_01: And I think that a lot of times companies are looking at and optimizing the
• 36:36
  SPEAKER_01: the recruiting process, but they tend to forget that it's a two-way street.
• 36:41
  SPEAKER_01: You know, the candidate wants to be able to evaluate the company too.
• 36:44
  SPEAKER_01: And as you said, it that tells you something about it.
• 36:46
  SPEAKER_01: Something about what kind of experience you might have as an employee there.
• 36:50
  SPEAKER_02: Yeah.
• 36:51
  SPEAKER_02: That's always been my thinking, right?
• 36:52
  SPEAKER_02: AI is meant to do all this optimization,
• 36:54
  SPEAKER_02: but at some point we need to stop and think what optimization are we trying to do
• 36:59
  SPEAKER_02: and what are the trade-offs that we're losing?
• 37:02
  SPEAKER_02: Because again,
• 37:04
  SPEAKER_02: for me as a candidate,
• 37:05
  SPEAKER_02: if you tell me my first interviews with AI,
• 37:07
  SPEAKER_02: I really am not going to come to that first interview.
• 37:09
  SPEAKER_02: I probably would try and figure out how I can send a model of myself to
• 37:13
  SPEAKER_02: the interview because why would I in my mind talk to an AI?
• 37:18
  SPEAKER_02: I've already filled out an application and I know
• 37:20
  SPEAKER_02: that application has probably been screened by AI.
• 37:24
  SPEAKER_02: Why would I record myself having a conversation?
• 37:27
  SPEAKER_02: Because again, I don't know what's going to happen with that data afterwards.
• 37:30
  SPEAKER_02: Are you going to use me?
• 37:31
  SPEAKER_02: Am I going to be your internal training candidate on here's an example of Oh,
• 37:36
  SPEAKER_02: this could go right or
• 37:37
  SPEAKER_02: Or wrong.
• 37:37
  SPEAKER_02: I don't know.
• 37:38
  SPEAKER_02: It's just very interesting.
• 37:40
  SPEAKER_01: So has any company's use of your data ever caused any problems for you?
• 37:45
  SPEAKER_01: Have you had any phishing attacks or any ways in which your privacy has been compromised?
• 37:49
  SPEAKER_01: Anything like that?
• 37:51
  SPEAKER_02: I feel like That's a new norm now.
• 37:54
  SPEAKER_02: I'm constantly getting data breaches it's to the point where
• 37:58
  SPEAKER_02: where I've locked all my credits with the different bureaus just
• 38:03
  SPEAKER_02: because every other day there's some extra data processor for a company
• 38:08
  SPEAKER_02: that I didn't know about.
• 38:09
  SPEAKER_02: So it's their third party data processor that had the breach.
• 38:13
  SPEAKER_02: And it's like, oh, two years ago you, this was your data.
• 38:17
  SPEAKER_02: Data.
• 38:17
  SPEAKER_02: And then here's a year of identity protection from the dark web and
• 38:22
  SPEAKER_02: and every few days are or every few weeks at one point,
• 38:25
  SPEAKER_02: I feel like I was getting a letter.
• 38:26
  SPEAKER_02: And so I was like, okay,
• 38:27
  SPEAKER_02: I need to just lock all my credit because
• 38:29
  SPEAKER_02: at this point it doesn't feel like it's safe out there.
• 38:33
  SPEAKER_02: It probably has been sold multiple times.
• 38:35
  SPEAKER_02: I don't know if these causes we're specifically using AI
• 38:39
  SPEAKER_02: Specifically because of AI, but again, I I don't know if it was.
• 38:44
  SPEAKER_02: Wasn't because a lot of companies will use third parties to process the data.
• 38:50
  SPEAKER_02: Those third parties might get compromised.
• 38:52
  SPEAKER_02: And how do you go?
• 38:53
  SPEAKER_02: Got compromised.
• 38:54
  SPEAKER_02: Usually they'll tell you or they don't tell you,
• 38:57
  SPEAKER_02: but yeah,
• 38:58
  SPEAKER_02: I feel like there's been a lot of,
• 38:59
  SPEAKER_02: at least in the last few years, a lot of, Hey, your data might've been compromised.
• 39:04
  SPEAKER_02: And it's been overall a lot of my previous companies,
• 39:08
  SPEAKER_02: it would be from maybe my a 4 -on -key provider.
• 39:12
  SPEAKER_02: Provider or my the HSA provider.
• 39:15
  SPEAKER_02: Provider.
• 39:15
  SPEAKER_02: So it was a lot of the banking institution providers that were getting compromised,
• 39:20
  SPEAKER_02: which was interesting.
• 39:21
  SPEAKER_02: But I've also had from previous phone companies that I've use or or some third party
• 39:29
  SPEAKER_02: of the third party of the third party was really weird how we got here,
• 39:33
  SPEAKER_02: but the data had been compromised.
• 39:35
  SPEAKER_02: And so now I get one year of protection for the dark web.
• 39:40
  SPEAKER_02: The internet, but after after that year, I can't get the new social or anything, right?
• 39:44
  SPEAKER_02: It's already out there.
• 39:45
  SPEAKER_02: These data compromises, I feel like, And I'll see you next time.
• 39:48
  SPEAKER_02: Definitely something you know, that that are happening a lot more.
• 39:53
  SPEAKER_02: I don't know if they're specifically from an AI perspective, but I don't know if they're not.
• 39:57
  SPEAKER_01: Yeah, I mean,
• 39:58
  SPEAKER_01: data brokers buying and selling our data and having data breaches has been going on
• 40:03
  SPEAKER_01: for many years, long before GEMGVI really became common.
• 40:06
  SPEAKER_01: It may give more of a financial incentive for them to do it because they have
• 40:10
  SPEAKER_01: a new market to sell to, right?
• 40:12
  SPEAKER_01: But I'm curious, do you ever take up the offers of the free protection?
• 40:16
  SPEAKER_02: I initially I used to do it.
• 40:19
  SPEAKER_02: It and then because it was happening so much,
• 40:25
  SPEAKER_02: at one point I think I had protection for like three years and
• 40:28
  SPEAKER_02: then another one happened and I was like, oh great, I have another extra year.
• 40:31
  SPEAKER_02: This one was just about to run out.
• 40:34
  SPEAKER_02: So I specifically just went and across all the bureaus, I just locked my credits.
• 40:39
  SPEAKER_02: So that makes me feel a lot safer.
• 40:41
  SPEAKER_02: I can't even open a credit card right now.
• 40:43
  SPEAKER_02: No one can, which is good.
• 40:45
  SPEAKER_02: And then I try and monitor where I'm using my credit cards
• 40:49
  SPEAKER_02: that I have opened right now So that if they do
• 40:51
  SPEAKER_02: do get compromised, at least I understand that, well, I did the best that I could.
• 40:56
  SPEAKER_02: And I do try to use them when I can,
• 40:58
  SPEAKER_02: but I don't know if they're actually doing anything for me because again,
• 41:01
  SPEAKER_02: if and if the data is on the dirt,
• 41:02
  SPEAKER_02: dark web or if it's been sold, after the year is over.
• 41:05
  SPEAKER_02: Up, the data is still out there.
• 41:08
  SPEAKER_02: Protections are only to monitor.
• 41:11
  SPEAKER_02: It's not necessarily going to prevent anything.
• 41:13
  SPEAKER_02: It's just, hey, we noticed your data was used, but it's not going to stop it.
• 41:18
  SPEAKER_01: Yeah.
• 41:18
  SPEAKER_01: I've gotten a lot of those data breach notifications myself, and I've gotten those offers.
• 41:22
  SPEAKER_01: And one thing my husband pointed out is that there are all these
• 41:26
  SPEAKER_01: and made up protection companies now.
• 41:29
  SPEAKER_01: How do you know you can trust them?
• 41:31
  SPEAKER_01: Them with the data that you
• 41:32
  SPEAKER_01: You need to give them in order for them to protect you.
• 41:37
  SPEAKER_01: So we don't take those up because, as you said.
• 41:40
  SPEAKER_01: Said, you have to give them all this private information in order for them
• 41:43
  SPEAKER_01: to protect your private information.
• 41:45
  SPEAKER_01: And how do you know they can meet you?
• 41:46
  SPEAKER_02: Be trusted yeah that's an interesting point as well and I think that's why I went
• 41:50
  SPEAKER_02: the extra step to lockdown or credit because
• 41:53
  SPEAKER_02: in my mind I'm sure they also will get compromised eventually are they going
• 41:57
  SPEAKER_02: to tell people when they get compromised and who's going to protect them
• 42:00
  SPEAKER_02: but yeah that's an interesting point because you definitely We have to give them
• 42:04
  SPEAKER_02: a lot of advice.
• 42:04
  SPEAKER_02: Lot of data just to get it started and then you're basically verifying who you are
• 42:09
  SPEAKER_02: so it's if they get compromised then I and I don't know, I don t know.
• 42:13
  SPEAKER_02: Know who to go to next Right.
• 42:16
  SPEAKER_01: Well, that's a question and then we can talk about anything else that you want.
• 42:19
  SPEAKER_01: We've noticed that public distrust of AI intelligence.
• 42:22
  SPEAKER_01: Tech companies has been growing and I think that's good reason
• 42:25
  SPEAKER_01: because we're finding out more about what they're doing with their data.
• 42:28
  SPEAKER_01: But do you think there is any company that you CanTrust?
• 42:33
  SPEAKER_01: Trust or would trust or is that not possible?
• 42:36
  SPEAKER_01: And if it's possible,
• 42:37
  SPEAKER_01: what would be the main thing they would have to do for you to trust one
• 42:42
  SPEAKER_01: of these companies with your data?
• 42:43
  SPEAKER_02: I think it's a deeper question.
• 42:46
  SPEAKER_02: For me, when it comes to that trust of your data, I trust, but I always verify.
• 42:54
  SPEAKER_02: And so I think we'd come companies right now,
• 42:59
  SPEAKER_02: the companies that are telling you, yes, Yes, this is her.
• 43:02
  SPEAKER_02: How we're going to handle your data.
• 43:04
  SPEAKER_02: This is how we handle our third parties.
• 43:08
  SPEAKER_02: It gives you a level of trust but at the same time a breach can happen and
• 43:14
  SPEAKER_02: and what happens with it.
• 43:15
  SPEAKER_02: With that breach which is...
• 43:18
  SPEAKER_02: how do they handle things where okay the data was taken do they have
• 43:23
  SPEAKER_02: a good track history of doing it it's one of those things I try to think of it.
• 43:30
  SPEAKER_02: Think about it not too much but also think about it because I think
• 43:35
  SPEAKER_02: with every company inherently, they want to do well with your data.
• 43:39
  SPEAKER_02: I don't think any company wakes up and they're like,
• 43:42
  SPEAKER_02: yeah, today we're going to take a bunch of data from our employees.
• 43:45
  SPEAKER_02: That's the plan and the mission for Q1.
• 43:47
  SPEAKER_02: I don't think there are any companies like that.
• 43:49
  SPEAKER_02: I think what tends to happen is that companies come in different shapes and different sizes.
• 43:57
  SPEAKER_02: Bigger companies have teams of teams that handle the data,
• 44:02
  SPEAKER_02: handle the legal side, handle how data should
• 44:05
  SPEAKER_02: Should happen, not happen, how long they should retain it.
• 44:08
  SPEAKER_02: Is it seven years?
• 44:09
  SPEAKER_02: Even from a CRM perspective, those standard operating procedures are written out.
• 44:15
  SPEAKER_02: We hope people are following them.
• 44:17
  SPEAKER_02: That's another thing.
• 44:18
  SPEAKER_02: We hope people are following them.
• 44:20
  SPEAKER_02: I think sometimes of smaller companies, things are just happening quicker.
• 44:23
  SPEAKER_02: They're more rapid to So, acceleration.
• 44:27
  SPEAKER_02: They just haven't gotten to that point in their journey yet where they're having
• 44:30
  SPEAKER_02: to think about it because maybe they started as a very small company.
• 44:35
  SPEAKER_02: But either way,
• 44:37
  SPEAKER_02: for me, I trust companies where I can clearly see this is how we handle your data.
• 44:44
  SPEAKER_02: This is what we're trying to do.
• 44:45
  SPEAKER_02: Trust is immediate.
• 44:47
  SPEAKER_02: It's one of the reasons I work in the Salesforce ecosystem,
• 44:50
  SPEAKER_02: because I know at least Salesforce as a company They have a whole page that I can go
• 44:55
  SPEAKER_02: to and I can say, yes, this is what's going to happen.
• 44:59
  SPEAKER_02: And there were some some partner companies that had gotten hacked.
• 45:05
  SPEAKER_02: It was all over the news.
• 45:07
  SPEAKER_02: Salesforce specifically said we're not going to pay a ransom for what happened over here.
• 45:13
  SPEAKER_02: Over here with that specific company.
• 45:15
  SPEAKER_02: I think it was Drift.
• 45:16
  SPEAKER_02: Drift had a hack recently and the hackers are trained.
• 45:19
  SPEAKER_02: Trying to get Salesforce to pay.
• 45:20
  SPEAKER_02: They're like, well, we're not going to negotiate with you.
• 45:23
  SPEAKER_02: And I think just having that stand, like, yes,
• 45:26
  SPEAKER_02: this is a lot of company data that has gotten taken,
• 45:29
  SPEAKER_02: but one of the things they also did was work with this company to, one,
• 45:33
  SPEAKER_02: delisted from the AppExchange so that no one else is downloading the app,
• 45:37
  SPEAKER_02: but then also work with their cybersecurity team
• 45:40
  SPEAKER_02: to figure out how we can prevent this from happening.
• 45:43
  SPEAKER_02: And there have been a lot more security updates that have happened now because
• 45:48
  SPEAKER_02: of our - all these hackings that have happened from not Salesforce themselves,
• 45:52
  SPEAKER_02: but companies that use Salesforce.
• 45:54
  SPEAKER_02: And so I think seeing things like that makes me feel,
• 45:58
  SPEAKER_02: for lack of better words, more trusting towards companies like that.
• 46:03
  SPEAKER_02: But other companies are doing it as well.
• 46:04
  SPEAKER_02: There are other big companies in the enterprise where an incident happened.
• 46:09
  SPEAKER_02: Happens.
• 46:10
  SPEAKER_02: How they deal with the incident is how I know,
• 46:12
  SPEAKER_02: okay, I can trust this company because data is going to get leaked.
• 46:17
  SPEAKER_02: We know that.
• 46:17
  SPEAKER_02: Hacks are going to happen.
• 46:20
  SPEAKER_02: Protections are going to be breached.
• 46:21
  SPEAKER_02: It happens.
• 46:22
  SPEAKER_02: But how they deal with the incident is what I'm looking for.
• 46:25
  SPEAKER_02: Are you giving a root cause analysis?
• 46:26
  SPEAKER_02: Are you saying, we know we messed up.
• 46:30
  SPEAKER_02: This is how we're going to put steps in place.
• 46:31
  SPEAKER_02: And then you can actually see these steps in place from a public perspective,
• 46:35
  SPEAKER_02: not a Yeah, we went behind closed doors.
• 46:37
  SPEAKER_02: We fixed it.
• 46:38
  SPEAKER_02: It's fixed.
• 46:39
  SPEAKER_02: And we don't know if it's fixed.
• 46:40
  SPEAKER_02: It's just, it's fixed.
• 46:41
  SPEAKER_02: We don't know.
• 46:42
  SPEAKER_02: What does that mean?
• 46:43
  SPEAKER_02: So to summarize everything, I naturally will trust.
• 46:48
  SPEAKER_02: I like to see what happens when an incident happens because then I know how
• 46:51
  SPEAKER_02: a company can handle it.
• 46:53
  SPEAKER_02: Companies come in different shapes and sizes.
• 46:55
  SPEAKER_02: And so bigger enterprises have more for teams of teams.
• 47:01
  SPEAKER_02: Teams that handle these data protections.
• 47:03
  SPEAKER_02: Smaller companies might not have gotten there yet in their journey,
• 47:06
  SPEAKER_02: but I think I hope inherently there are no companies that are saying like today's
• 47:11
  SPEAKER_02: the day we're going to just leak some data and see what happens.
• 47:14
  SPEAKER_01: All right.
• 47:15
  SPEAKER_01: Fair enough.
• 47:16
  SPEAKER_01: Well, thank you for joining me and sharing your thoughts on these questions.
• 47:20
  SPEAKER_01: Is there anything else that you said?
• 47:21
  SPEAKER_01: That you would like to say?
• 47:22
  SPEAKER_01: For instance, I have Simmons idea.
• 47:23
  SPEAKER_01: Information that you're working on a chapter in a book.
• 47:27
  SPEAKER_02: Yes.
• 47:28
  SPEAKER_02: I will say that But right now I'm I'm very passionate about AI.
• 47:33
  SPEAKER_02: I'm very passionate about how it's affecting our resources from
• 47:40
  SPEAKER_02: a data center perspective, from a electricity per se.
• 47:43
  SPEAKER_02: Perspective, and from a water source per perspective.
• 47:47
  SPEAKER_02: Our volume two book will be out soon so that you can read all about my perspective on that
• 47:55
  SPEAKER_02: Different side of, yes,
• 47:57
  SPEAKER_02: we should use AI,
• 47:58
  SPEAKER_02: but also we should understand how our resources are being affected.
• 48:03
  SPEAKER_02: You can find me at techsips.nadinalisbin.com.
• 48:07
  SPEAKER_02: It's on Substacks.
• 48:09
  SPEAKER_02: Or if you to literally search.
• 48:11
  SPEAKER_02: Search TechSips with Nadina, it will be at least the first service.
• 48:15
  SPEAKER_01: Search that comes up.
• 48:17
  SPEAKER_01: Thank you so much, Nadina.
• 48:18
  SPEAKER_01: It was great to talk with you today.
• 48:19
  SPEAKER_01: I appreciate you making time for the interview.
• 48:21
  SPEAKER_02: Thank you so much, Karen.
• 48:23
  SPEAKER_02: It was great talking with you as well.
• 48:26
  SPEAKER_00: About this interview series and newsletter.
• 48:28
  SPEAKER_00: This post is part of our AI 6P interview series on AI software and webware.
• 48:36
  SPEAKER_00: It showcases how real people around the world are using their wetware, brains,
• 48:40
  SPEAKER_00: and human intelligence with AI-based software tools who are being affected by AI.
• 48:46
  SPEAKER_00: In AI Pods And we're all being affected by AI nowadays in our daily lives,
• 48:51
  SPEAKER_00: perhaps more than we realize.
• 48:53
  SPEAKER_00: For some examples, see post but I don't use AI.
• 48:57
  SPEAKER_00: Thank you very much for joining us today.
• 48:58
  SPEAKER_00: We want to hear from a diverse pool of people worldwide in a variety of roles.
• 49:03
  SPEAKER_00: If you're interested in being a featured interview guest,
• 49:06
  SPEAKER_00: anonymous or with credit, please get in touch.
• 49:12
  SPEAKER_00: 6 6P's in AI Pods AI6P is a 100 % reader support.
• 49:17
  SPEAKER_00: Reader-supported publication.
• 49:19
  SPEAKER_00: All new posts are free to read and listen to.
• 49:23
  SPEAKER_00: To automatically receive new AI6P posts and support our work, consider becoming a subscriber.
• 49:29
  SPEAKER_00: It's free.