Latest / Humanity Working / The Truth About AI Skills

Transcript

The Truth About AI Skills · Humanity Working

• 0:05
  SPEAKER_01: All right, Paul, let's talk AI upskilling.
• 0:07
  SPEAKER_01: Everyone's throwing that term around, but what does it actually mean?
• 0:11
  SPEAKER_00: Yeah.
• 0:13
  SPEAKER_00: Yeah, let's talk AI upskilling.
• 0:15
  SPEAKER_00: I'm really glad we are, actually.
• 0:16
  SPEAKER_00: It's been something that's been obsessing me, I think, for the for the last two to three years.
• 0:23
  SPEAKER_00: Um, I mean, like basic definition, giving people the skills they need to thrive alongside AI, but there's so much, you know, I can say that in one sentence, but there's so much that goes underneath that, you know.
• 0:37
  SPEAKER_01: Yeah, I think uh the way that we've talked about it before, and the way that way it seems to me is is, you know, do we have a new do we have a new team member or are we using a tool?
• 0:48
  SPEAKER_01: You know, and I think that that's a really important district distinction.
• 0:51
  SPEAKER_00: Yeah, I I completely agree, actually.
• 0:53
  SPEAKER_00: I mean, this is the this is the thing I think that almost all traditional AI upskilling gets wrong.
• 1:00
  SPEAKER_00: You know what it does?
• 1:01
  SPEAKER_00: It starts from the premise of the of the engineering in a sense, right?
• 1:05
  SPEAKER_00: Um, so what happens is that you're treating it as almost like a traditional piece of technology, and you're saying, okay, if you're an AI developer, then you need to understand this technology at a 400 level, right?
• 1:23
  SPEAKER_00: And then you kind of work your way down and you go, okay, what's the slightly simpler version of that at a 300 level?
• 1:30
  SPEAKER_00: What's the slightly simpler version of that at a 200 level?
• 1:32
  SPEAKER_00: And what's the really simple version of that at a 100 level?
• 1:36
  SPEAKER_00: And because it starts with the idea of 400 and then simplifying down to 100, where do you end up?
• 1:43
  SPEAKER_00: Well, where you end up with is like, okay, well, here's how to prompt engineer, right?
• 1:49
  SPEAKER_00: Because that's the it's an interface.
• 1:51
  SPEAKER_00: Here's how you interact with this interface.
• 1:54
  SPEAKER_00: Very traditional.
• 1:55
  SPEAKER_00: It would almost be like you were uh learning Excel.
• 1:59
  SPEAKER_00: It's it's sort of like the version of that, right?
• 2:02
  SPEAKER_00: Because if you if you're in Excel, here are cells, here's a row, here are columns, here's how you use it, here's how you, you know, et cetera, et cetera.
• 2:09
  SPEAKER_00: But AI is totally different, right?
• 2:12
  SPEAKER_00: AI changes how we work.
• 2:16
  SPEAKER_00: AI is unlike, uh particularly generative AI, is unlike any other technology really that we've ever worked with.
• 2:23
  SPEAKER_00: The closest I would say that AI is to in terms of something that we've ever worked with previously would be something like, I don't know, Siri or Alexa or uh or even a search, the search bar in Google.
• 2:35
  SPEAKER_00: But it's not like any other sort of traditional technology that we would use.
• 2:40
  SPEAKER_00: And if it's not changing the way we work, we're not working with it right.
• 2:45
  SPEAKER_00: And so your example of it being a team member is perfect because Ethan Moloch uses this phrase, right?
• 2:51
  SPEAKER_00: AI is like an intern, a hardworking intern that lies a bit.
• 2:54
  SPEAKER_01: Yeah.
• 2:55
  SPEAKER_00: And it is a um, that's how you should think about it.
• 2:58
  SPEAKER_00: And so you would go through a completely different type of training if you said, we're bringing a new person in, they work really hard, they they kind of lie a bit occasionally, and you need to figure out how to work with them.
• 3:12
  SPEAKER_00: That's a totally different training course than a prompt engineering course.
• 3:16
  SPEAKER_01: Yeah, yeah, because almost prompt engineering isn't an isn't isn't nearly enough because you could say, here are the 25 prompts you need to work with AI, and you would not only be limiting yourself, but also boxing in the tool to the point where it doesn't help you that much, right?
• 3:33
  SPEAKER_00: Well, yeah, not only that, but the tool itself is so different, right?
• 3:37
  SPEAKER_00: I mean, when ChatGPT launched, nine months later, half of the knowledge workers in the US were using its fastest adopted piece of technology in the history of technology because it was so radically different, right?
• 3:50
  SPEAKER_00: And this was without companies asking people to do it or anything like that, right?
• 3:53
  SPEAKER_00: So people intrinsically recognized that it was unique and that it was it was changing the way that they worked.
• 3:59
  SPEAKER_00: But it is, but because it's so different to your to your point, framing the way in which you work with it in that way is limiting, but it itself has changed massively over that period of time.
• 4:13
  SPEAKER_00: Just in GPT itself, we have like five different mod five different active models right now in GPT, right?
• 4:19
  SPEAKER_00: I don't know anybody who really knows how to use one or the other, or um, you know, other than people that are going through our programs, of course, that's very different.
• 4:26
  SPEAKER_00: But but hardly anybody knows how to choose one ver uh one over another, yeah.
• 4:33
  SPEAKER_00: Uh what the likelihood, how to deal with it when it is getting things wrong, how to understand that it is, you know, all of those things.
• 4:40
  SPEAKER_00: They're just not uh they're not figuring that out.
• 4:42
  SPEAKER_00: And a prompt engineering course isn't gonna do any of that.
• 4:44
  SPEAKER_00: Yeah, it's about I don't know, five percent of what people should be focusing on.
• 4:49
  SPEAKER_01: Yeah.
• 4:50
  SPEAKER_01: It's interesting because it feels like they, when you use different ones, they have almost different personalities.
• 4:56
  SPEAKER_01: So like again, again, it's like, oh, they're they're they're engaging a little bit differently, you know.
• 5:01
  SPEAKER_01: And and I think it's funny too when you think about like you'd all see these videos when Alexa and Siri started coming out, videos on YouTube of people just like literally being so mean to their Alexa, or you know, it's or some people are super polite, and you're just like, you don't have to be that night, you know what I mean?
• 5:17
  SPEAKER_01: So, you know, and how to we struggled with like getting it to hear us the right way.
• 5:22
  SPEAKER_01: That's an old joke too, now is like yeah, you say something and it tells you a different thing.
• 5:26
  SPEAKER_01: There's a great Alpha Romeo commercial where the guy is uh sits, it's like set in the future and he sits down and it's a self-driving car, but you have to tell it what to do.
• 5:36
  SPEAKER_01: And it's like, take me to work.
• 5:38
  SPEAKER_01: And it's like, I'll turn on the climate control.
• 5:40
  SPEAKER_01: And he's just getting so frustrated, and eventually he's just like, I just want to drive.
• 5:44
  SPEAKER_01: And he like snaps into you know, an Alpha Romeo, like this might be the last time you get to drive, kind of thing.
• 5:49
  SPEAKER_01: But it's like the the same joke we saw about like, oh, it doesn't understand me when I talk to it, that's happening now too.
• 5:57
  SPEAKER_01: It's just on such a bigger level, and you're typing these prompts, and it's like, no, you got to figure out how to add this companion, you know.
• 6:07
  SPEAKER_01: So, like, what's what's the big message then for companies that are trying to figure out what to do?
• 6:14
  SPEAKER_00: Well, there's several, but I mean, some of it we've touched, we've touched on already, um, which is that it's sort of human but not, and understanding that um gives us clues as to the kind of skills that you need that that you need to be able to build.
• 6:33
  SPEAKER_00: So, yes, hardworking internet lies a bit, great mental model.
• 6:37
  SPEAKER_00: Um, but also related to that is Matt, the vast majority of generative AI is massively overconfident, right?
• 6:43
  SPEAKER_00: So, for example, it never says has has GPT ever said to you said, Matt, I don't have a clue what you're talking about, right?
• 6:52
  SPEAKER_00: No, you you tell it some half-baked piece of information and it goes, okay, you know.
• 6:58
  SPEAKER_01: I can find I can find stuff that relates to that.
• 7:00
  SPEAKER_01: Yeah.
• 7:01
  SPEAKER_00: Yeah, exactly.
• 7:02
  SPEAKER_00: And so it pretends that it understands perfect perfectly what you what you need to know.
• 7:06
  SPEAKER_00: So it's also dealing with its overconfidence and understanding its overconfidence um and interacting with it.
• 7:12
  SPEAKER_00: But then also, of course, figuring out how to use it on its own in a team setting.
• 7:16
  SPEAKER_00: We're starting to see emerging right now, sort of like almost like AI being connected to other AI and different agents working with each other and the amplification effect associated with that, figuring out how to work with that.
• 7:28
  SPEAKER_00: But ultimately it comes down very simply to you need to be able to use it safely and effectively, um, in a way that makes work better as well as you know, the output of work better, not just faster and uh and more efficient.
• 7:46
  SPEAKER_00: Um, and you need to make sure that you're minimizing the risk to yourself, to your team, to your organization.
• 7:53
  SPEAKER_00: And the key thing is, the most important thing is that organizations that think this is a skill set, that basic skill set that I just talked about is a skill set for a subset of the company and not the whole company, are massively missing a trick.
• 8:08
  SPEAKER_00: Right?
• 8:09
  SPEAKER_00: One human misusing AI in the organization, unintentionally in many cases, one human can cause significant damage to the organization.
• 8:18
  SPEAKER_00: Um, and so there is a base level of AI literacy centered around these principles, not just prompt engineering, that needs to be in place in every organization.
• 8:31
  SPEAKER_00: And companies that are not thinking about that and actually acting on it right now are missing a trick and it's not a one and done because AI itself is moving so fast.
• 8:41
  SPEAKER_01: And I would add to that also don't think about it like, oh, I have AI now, what can I replace in my organization?
• 8:50
  SPEAKER_01: But more like if I can put this in everyone's hands and they know what to do with it, I'm exponentially expanding my capability as opposed to you know to the infinite growth point.
• 9:03
  SPEAKER_01: Is like we're not trying to necessarily grow like that.
• 9:11
  SPEAKER_01: Maybe flat.
• 9:12
  SPEAKER_01: Now we can stay flat and improve our growth as an organization because of this, not just how many people can this replace, right?
• 9:22
  SPEAKER_01: Because precisely, yeah.
• 9:24
  SPEAKER_01: It's the wrong answer, right?
• 9:25
  SPEAKER_01: Yeah.
• 9:26
  SPEAKER_00: Yeah.
• 9:26
  SPEAKER_01: It's very interesting.
• 9:27
  SPEAKER_00: How am I going to construct my teams?
• 9:29
  SPEAKER_00: How am I going to construct my organization?
• 9:31
  SPEAKER_00: How are they going to work with AI?
• 9:33
  SPEAKER_00: Um, all of those questions are questions that most organizations are scratching the surface on, and they don't have basic AI literacy in their organizations to even build from.
• 9:45
  SPEAKER_01: Well, gotta fix that, I suppose.
• 9:47
  SPEAKER_01: Let's do it.
• 9:48
  SPEAKER_01: All right.
• 9:49
  SPEAKER_01: Thanks, Paul.
• 9:50
  SPEAKER_01: Thank you.
• 9:51
  SPEAKER_01: Humanity Working is produced and edited for Billion Minds by Matt Neal.
• 9:54
  SPEAKER_01: Our interviewer is Paul Slater.
• 9:56
  SPEAKER_01: If you like this podcast, please subscribe, like, and recommend us to others.
• 10:00
  SPEAKER_01: It helps us reach more people.
• 10:02
  SPEAKER_01: If you want to learn more about how Billion Minds gets companies ready for the future of work by building adaptable, resilient employees, look for them on LinkedIn or visit their website at billionminds.com.