Latest / Connor with Honor: AI, Real Estate and Real Life / The Human Transition: AI Needs a Plan for People

Transcript

The Human Transition: AI Needs a Plan for People · Connor with Honor: AI, Real Estate and Real Life

• 0:01
  SPEAKER_00: What happens when the most capable intelligence in the room is something that we built?
• 0:06
  SPEAKER_00: Now suppose it can solve problems we can't solve, see connections that we keep missing and do work faster than we can follow.
• 0:14
  SPEAKER_00: What happens to our place in that room?
• 0:16
  SPEAKER_00: And it's a serious question, but it's also very exciting.
• 0:20
  SPEAKER_00: Because if we handle this well and do it right, greater capability could be more people are getting a real shot at a better life.
• 0:27
  SPEAKER_00: And that part we need to start building right now.
• 0:30
  SPEAKER_00: That part is the human transition.
• 0:33
  SPEAKER_00: Let me explain language I'm going to use.
• 0:36
  SPEAKER_00: Personally, I prefer synthetic intelligence, synthetic general intelligence, and synthetic superintelligence.
• 0:43
  SPEAKER_00: That's my framing, rather than an official new scientific classification.
• 0:47
  SPEAKER_00: I'm sure I've heard this mentioned somewhere, so credit.
• 0:50
  SPEAKER_00: Well, if you're out there, thank you.
• 0:53
  SPEAKER_00: Synthetic means engineered in this conversation.
• 0:57
  SPEAKER_00: I'm talking about intelligence created through technology.
• 0:59
  SPEAKER_00: I'm not claiming that it's duplicate of a human brain or it's a machine producing a convincing answer, having feelings.
• 1:07
  SPEAKER_00: The word helps me keep something important in my view.
• 1:10
  SPEAKER_00: People are building these systems.
• 1:13
  SPEAKER_00: People were the nexus of these systems being built.
• 1:17
  SPEAKER_00: And choosing where they go, having chosen where they are going and deciding how much authority they receive.
• 1:24
  SPEAKER_00: And that's right now.
• 1:26
  SPEAKER_00: And we are about the middle of October 2026.
• 1:29
  SPEAKER_00: So start with what's here.
• 1:31
  SPEAKER_00: So today's systems can write, translate, analyze information, create images, and build software.
• 1:36
  SPEAKER_00: Their usefulness depends on the system, the task, the information available, and how somebody checks the result.
• 1:42
  SPEAKER_00: They can also deliver a wrong answer with remarkable confidence.
• 1:46
  SPEAKER_00: A beautiful explanation can contain a made-up fact.
• 1:49
  SPEAKER_00: A working demonstration can fail when the situation changes.
• 1:53
  SPEAKER_00: We need room for genuine amazement and the very ordinary question, how do we know this is right?
• 2:00
  SPEAKER_00: Then widen the picture.
• 2:01
  SPEAKER_00: Synthetic general intelligence is how all refer to the idea usually called artificial general intelligence, broadly capable systems that could learn and perform across a wide range of cognitive tasks.
• 2:15
  SPEAKER_00: People disagree about the exact definition and the test for getting there.
• 2:20
  SPEAKER_00: A system doing one thing better than us doesn't settle that debate.
• 2:24
  SPEAKER_00: Even doing several things impressively leaves questions about reliability, unfamiliar situations, and how much help it needed.
• 2:32
  SPEAKER_00: Those distinctions matter when somebody announces that everything has changed.
• 2:37
  SPEAKER_00: Let's take that further still.
• 2:39
  SPEAKER_00: Synthetic superintelligence describes a hypothetical level of capability substantially beyond human capability across many important intellectual activities.
• 2:50
  SPEAKER_00: Now that's a possibility to examine carefully with no settled arrival date and no guarantee about the path.
• 2:57
  SPEAKER_00: Greater capability wouldn't automatically establish wisdom, human values, or the right to make decisions for us.
• 3:05
  SPEAKER_00: And the uncertainty gives us a reason to prepare thoughtfully.
• 3:08
  SPEAKER_00: We can improve how we handle today's systems while studying what would be necessary if much more powerful systems emerge.
• 3:16
  SPEAKER_00: Now here's where my optimism gets practical.
• 3:19
  SPEAKER_00: I think, I think we need a transition moonshot.
• 3:23
  SPEAKER_00: A serious effort to help people understand shape and benefit from this change.
• 3:29
  SPEAKER_00: The technical work is receiving enormous attention.
• 3:32
  SPEAKER_00: I'd like the human preparation to receive comparable seriousness.
• 3:37
  SPEAKER_00: Who gets the training?
• 3:38
  SPEAKER_00: Who has a stay, a say?
• 3:40
  SPEAKER_00: What happens when a job changes?
• 3:42
  SPEAKER_00: How do we recognize harm early?
• 3:45
  SPEAKER_00: How do we distribute the benefits?
• 3:47
  SPEAKER_00: Those questions belong inside the project from the beginning with budgets, people responsible, and results we can actually inspect.
• 3:55
  SPEAKER_00: The moon landing analogy has limits.
• 3:58
  SPEAKER_00: A society is messier than a single engineering mission.
• 4:02
  SPEAKER_00: With different communities, we'll choose different approaches.
• 4:05
  SPEAKER_00: But the useful idea is shared purpose backed by sustained work.
• 4:09
  SPEAKER_00: Saying that we care about the future?
• 4:11
  SPEAKER_00: Well, that's easy.
• 4:12
  SPEAKER_00: Building the training, running the experiments, fixing what fails, and sticking with people, human beings, so nobody gets left behind, nobody gets screwed over through the transition.
• 4:24
  SPEAKER_00: That takes commitment.
• 4:25
  SPEAKER_00: Now, human transition deserves that kind of ambition.
• 4:28
  SPEAKER_00: We should be able to point to something more than more substantial than a conference slogan.
• 4:34
  SPEAKER_00: Human involvement starts long before somebody types a question.
• 4:39
  SPEAKER_00: People choose what to build, what to test, what to connect, and where to deploy it.
• 4:44
  SPEAKER_00: Some systems can already carry out sequences of actions using tools and delegated permissions without somebody approving every step, and that's just going to get even more and more we we hate lifting heavy things for the most part.
• 4:57
  SPEAKER_00: And if we can have something else do it, well, we might not pay that much attention.
• 5:02
  SPEAKER_00: That does create real possibilities and also real failure modes.
• 5:06
  SPEAKER_00: The important question becomes what authority did we give this system and what happens if it uses that authority badly?
• 5:13
  SPEAKER_00: Responsibility needs to follow the people and organizations with the power to make those choices.
• 5:19
  SPEAKER_00: My investigators' instinct, as an LAPD cop for a long time, that is to slow down at the confident claim.
• 5:28
  SPEAKER_00: Now, what's the evidence?
• 5:29
  SPEAKER_00: Where did it come from?
• 5:31
  SPEAKER_00: What would contradict it?
• 5:32
  SPEAKER_00: What benefits if everybody believes it?
• 5:35
  SPEAKER_00: Those are useful questions for a product demonstration.
• 5:38
  SPEAKER_00: A frightening prediction or a promise that your business will magically run itself.
• 5:42
  SPEAKER_00: Ask to see the boring examples.
• 5:44
  SPEAKER_00: Ask what happened when the input was incomplete.
• 5:47
  SPEAKER_00: Ask how errors were discovered.
• 5:49
  SPEAKER_00: If we're going to trust a system with consequential work, we should be curious about where it screwed up.
• 5:55
  SPEAKER_00: Picture human knowledge is a pyramid built over generations.
• 5:59
  SPEAKER_00: Every useful discovery gives somebody else a place to stand.
• 6:03
  SPEAKER_00: A teacher helps a student understanding or understand something.
• 6:07
  SPEAKER_00: That student improves a tool.
• 6:09
  SPEAKER_00: Somebody uses that tool to ask a better question.
• 6:12
  SPEAKER_00: Synthetic intelligence could help more people reach more of that accumulated knowledge, connect ideas and test possibilities.
• 6:19
  SPEAKER_00: That's an extraordinary prospect.
• 6:22
  SPEAKER_00: Imagine someone with a good idea and limited resources getting help to build an early version that they can actually show another human being.
• 6:29
  SPEAKER_00: There's a few billion folks in the world, several billion, as a matter of fact.
• 6:33
  SPEAKER_00: Imagine them all having access.
• 6:35
  SPEAKER_00: What would become of this world in a good way?
• 6:38
  SPEAKER_00: And potentially.
• 6:39
  SPEAKER_00: Well, that pyramid definitely needs solid foundations.
• 6:42
  SPEAKER_00: A mistaken claim repeated through a thousand polished summaries is still a mistaken claim.
• 6:48
  SPEAKER_00: Speak and spread errors as readily as insight, so the exciting version includes reliable sources, experiments, correction, and people who know enough to notice when something doesn't add up.
• 6:58
  SPEAKER_00: And yes, it's probably the boring work.
• 7:01
  SPEAKER_00: But we should teach those habits alongside the tools.
• 7:05
  SPEAKER_00: We don't have to teach a hammer, but a hammer does one thing and it requires a human to operate it.
• 7:10
  SPEAKER_00: This is different.
• 7:12
  SPEAKER_00: Now knowing how to ask for an answer is useful.
• 7:15
  SPEAKER_00: Knowing how to investigate an answer makes the usefulness much more durable, especially when the answer sounds exactly like what you wanted.
• 7:23
  SPEAKER_00: Imagine a small business here in Santa Clarita.
• 7:26
  SPEAKER_00: This is a hypothetical example.
• 7:27
  SPEAKER_00: It's staff spends hours turning customer requests into drafts, schedules, and follow-up messages.
• 7:34
  SPEAKER_00: They test a synthetic intelligence system on one narrow part of that work.
• 7:38
  SPEAKER_00: It prepares a draft using approved information, and a person checks it before anything goes out.
• 7:44
  SPEAKER_00: Maybe that helps.
• 7:45
  SPEAKER_00: Maybe the review takes longer than expected.
• 7:48
  SPEAKER_00: Either result teaches them something.
• 7:51
  SPEAKER_00: The experiment should tell them whether customers are better served and whether the people doing the work are actually better off.
• 7:58
  SPEAKER_00: Now bring the employees into the design.
• 8:01
  SPEAKER_00: A person answering the phone knows which requests are confusing.
• 8:05
  SPEAKER_00: The person fixing mistakes knows where a process breaks.
• 8:09
  SPEAKER_00: Their experience can tell you what a demonstration missed.
• 8:12
  SPEAKER_00: Give those people time to test, permission to question, and a way to report problems without being treated as obstacles.
• 8:18
  SPEAKER_00: If the system frees up time, then discuss what happens to that time.
• 8:23
  SPEAKER_00: We're going to just stare out the window like in school, or are we going to do something else?
• 8:26
  SPEAKER_00: Better customer service, maybe more skilled work, a more manageable day.
• 8:31
  SPEAKER_00: These are decisions that a business can make deliberately.
• 8:34
  SPEAKER_00: The employment question deserves that level of detail.
• 8:37
  SPEAKER_00: The International Labor Organization's research distinguishes exposure to generative technology from actual job losses.
• 8:44
  SPEAKER_00: Jobs contain different tasks, and its research points to transformation as a more likely overall outcome than wholesale replacement.
• 8:53
  SPEAKER_00: Now that doesn't promise a painless outcome for any particular work or business.
• 8:57
  SPEAKER_00: Some roles could shrink, some could change substantially, and opportunities could emerge elsewhere.
• 9:03
  SPEAKER_00: We should examine what actually happens, including who has access to those opportunities and whose face is the cost of getting there.
• 9:12
  SPEAKER_00: Consider the person whose role changes faster than their options do.
• 9:16
  SPEAKER_00: Imagine they're good at their job, they're good at their work, supporting a household, suddenly being told to reinvent themselves.
• 9:23
  SPEAKER_00: A link to an online course might help, but it doesn't create time, replace lost pay, or guaranteed an employer will value that new skill.
• 9:31
  SPEAKER_00: Now that's where a transition plan needs to become concrete, something that's adopted, something that's mass used, something that's believed in, something that becomes the focal point.
• 9:40
  SPEAKER_00: What training leads to an actual opening?
• 9:43
  SPEAKER_00: Who pays while they learn?
• 9:44
  SPEAKER_00: What support is available if that first plan doesn't work?
• 9:48
  SPEAKER_00: Paid learning belongs in this conversation.
• 9:51
  SPEAKER_00: One proposal is protected learning time during the work week connected to tasks people can practice with support.
• 9:58
  SPEAKER_00: Employers, training providers, and public institutions could test different ways to fund it.
• 10:04
  SPEAKER_00: We should judge those programs by whether people gain useful skill and real options with follow-up after the certificate.
• 10:11
  SPEAKER_00: A person ought to be able to ask, what can I do at the end of this that I couldn't do before?
• 10:17
  SPEAKER_00: And where can I use it?
• 10:18
  SPEAKER_00: Now keep an eye on the first rung of the career ladder also.
• 10:22
  SPEAKER_00: Suppose a company automates the simpler assignments where beginners used to learn.
• 10:26
  SPEAKER_00: How does somebody become experienced?
• 10:29
  SPEAKER_00: That's a design problem worth solving early as well.
• 10:32
  SPEAKER_00: Teams could build supervised apprenticeships around checking outputs, understanding consumers, customers, and of course working through difficult cases with an experienced colleague.
• 10:42
  SPEAKER_00: Whether those approaches work needs testing, well, the goal is to preserve a route into competence.
• 10:50
  SPEAKER_00: So young workers and people changing careers and old workers can develop judgment through real practice.
• 10:58
  SPEAKER_00: Follow the benefits all the way through.
• 11:00
  SPEAKER_00: If a team produces the more useful work in fewer hours, where does the gain go?
• 11:04
  SPEAKER_00: Some needs to cover cost and keep the organization healthy.
• 11:08
  SPEAKER_00: Beyond that, though, there are choices: higher pay, better service, investment and trading, profit sharing, or fewer hours where that makes sense.
• 11:16
  SPEAKER_00: Each option has trade-offs, and I like those choices discussed openly with the people whose work helped create the improvement.
• 11:24
  SPEAKER_00: Shared prosperity takes decisions about distribution alongside decisions about production.
• 11:31
  SPEAKER_00: Public policy belongs here as well.
• 11:33
  SPEAKER_00: Communities could compare trading support, help for people between jobs incentives and responsible adoption and ways to keep small businesses competitive.
• 11:41
  SPEAKER_00: Some people will argue for broader income support.
• 11:44
  SPEAKER_00: Others are going to favor different approaches.
• 11:46
  SPEAKER_00: We should look at costs, incentives, evidence, and the experience of the people affected.
• 11:51
  SPEAKER_00: No policy label is going to guarantee success.
• 11:54
  SPEAKER_00: A serious moonshot can include disagreement and still insist on a basic standard.
• 12:00
  SPEAKER_00: The people carrying the transition need practical options while it's happening.
• 12:05
  SPEAKER_00: Public understanding needs investment.
• 12:07
  SPEAKER_00: Libraries, community colleges, employers, and local groups could offer hands-on sessions where people try these tools and learn their limits.
• 12:15
  SPEAKER_00: Let somebody see a useful result and examine a plausible mistake.
• 12:19
  SPEAKER_00: Let what information is appropriate to share and how to check the service privacy settings.
• 12:24
  SPEAKER_00: Teach people to ask whether they're interacting with a person or a system.
• 12:28
  SPEAKER_00: Give them enough understanding to persist to participate in decisions.
• 12:32
  SPEAKER_00: People deserve a way in this conversation that doesn't require becoming a computer scientist.
• 12:38
  SPEAKER_00: The world includes people with unreliable connections, limited money, disabilities, and languages that technology serves unevenly.
• 12:46
  SPEAKER_00: A global transition has to make room for them.
• 12:49
  SPEAKER_00: That could mean accessible interfaces, local language education, shared access to tools and programs designed with the communities using them.
• 12:57
  SPEAKER_00: We could ask who can benefit without buying the most expensive subscription or living near a technology center.
• 13:04
  SPEAKER_00: A moonshot measured only by what the best equipped people can accomplish would leave a great deal of human potential untouched.
• 13:11
  SPEAKER_00: Here's a practical starting point.
• 13:13
  SPEAKER_00: A local pilot with a clear beginning and a public account of what happened.
• 13:17
  SPEAKER_00: Bring together willing employers, workers, educators, and community representatives.
• 13:23
  SPEAKER_00: Choose a few bounded uses with manageable consequences.
• 13:26
  SPEAKER_00: Agree on the baseline, the protections, and what would make the pilot stop.
• 13:31
  SPEAKER_00: Pay people for the learning involved.
• 13:33
  SPEAKER_00: Record the actual costs, and at the end share enough evidence to help others judge the results while protecting private information.
• 13:41
  SPEAKER_00: Then improve the approach before expanding it.
• 13:43
  SPEAKER_00: Measure what people experience.
• 13:46
  SPEAKER_00: How many errors occurred?
• 13:47
  SPEAKER_00: How much time went into checking the correction?
• 13:50
  SPEAKER_00: Did customers get better answers?
• 13:52
  SPEAKER_00: Did workers gain skills, earn more, or experience more pressure?
• 13:57
  SPEAKER_00: Do the benefits continue after the novelty wore off, including the cost of software training, supervision, and mistakes?
• 14:04
  SPEAKER_00: A first first draft, a fast first draft can be useful, but the full process determines the value.
• 14:11
  SPEAKER_00: If the numbers look good while the people doing the work describe a problem, investigate that difference instead of smoothing it away.
• 14:19
  SPEAKER_00: Basic human rights belong at the center.
• 14:22
  SPEAKER_00: Privacy, dignity, fair treatment, and the ability to challenge consequential decisions should shape how systems are used.
• 14:29
  SPEAKER_00: I want a person affected by an automated decision to know where to take a question and how to get a meaningful review.
• 14:35
  SPEAKER_00: The details are going to vary across institutions and countries.
• 14:39
  SPEAKER_00: The principle is that greater technical power should come with serious protection for the people it affects, including people with very little power to push back on their own.
• 14:50
  SPEAKER_00: Accountability needs actual names.
• 14:52
  SPEAKER_00: Who approved this use?
• 14:54
  SPEAKER_00: Who checks it?
• 14:55
  SPEAKER_00: Human being.
• 14:57
  SPEAKER_00: Who can suspend it?
• 14:58
  SPEAKER_00: Who investigates an incident that helps repair the harm?
• 15:02
  SPEAKER_00: Developers, vendors, deploying organizations, and users have different responsibilities according to their roles and control.
• 15:10
  SPEAKER_00: Saying the computer did it doesn't resolve those questions.
• 15:13
  SPEAKER_00: Keep records that allow an appropriate reviewer to reconstruct what happened with sensible limits on the private data collected.
• 15:20
  SPEAKER_00: Make it possible to learn from an incident without turning every person involved into a permanent surveillance target.
• 15:28
  SPEAKER_00: Security makes this practical.
• 15:30
  SPEAKER_00: Give a system only the access it needs for the task.
• 15:33
  SPEAKER_00: Separate drafting a message from sending one.
• 15:37
  SPEAKER_00: Suggesting a purchase from spending money from analyzing records from changing them.
• 15:42
  SPEAKER_00: Test what happens when it encounters misleading instructions or unfamiliar situations.
• 15:47
  SPEAKER_00: Set limits on the actions it can take and make access revocable.
• 15:51
  SPEAKER_00: As capabilities and autonomy increase, a small mistake can travel further before somebody notices.
• 15:57
  SPEAKER_00: The safeguard needs to match that reach, and the extra caution where consequences are difficult to reverse.
• 16:03
  SPEAKER_00: Human oversight needs substance.
• 16:05
  SPEAKER_00: Imagine a reviewer presented with hundreds of machine generated questions and almost no time to inspect them.
• 16:12
  SPEAKER_00: A person is technically involved, but how much judgment can they exercise?
• 16:17
  SPEAKER_00: Effective oversight needs expertise, time, information, and genuine authority to disagree.
• 16:23
  SPEAKER_00: Now, in some settings, keeping a task manual may be that sensible choice until the evidence improves.
• 16:29
  SPEAKER_00: Human choice should survive that arrival of a faster option.
• 16:33
  SPEAKER_00: We ought to be able to ask where automation helps and where we still want direct human involvement.
• 16:40
  SPEAKER_00: Progress can earn trust through visible evidence.
• 16:43
  SPEAKER_00: Start with limited use, test it under realistic conditions.
• 16:46
  SPEAKER_00: Correct what fails and expand when the results justify it.
• 16:50
  SPEAKER_00: Some experiments will disappoint.
• 16:52
  SPEAKER_00: We should be able to acknowledge that without treating the entire field as a failure.
• 16:56
  SPEAKER_00: Some will genuinely be impressive.
• 16:59
  SPEAKER_00: We should be able to celebrate them without pretending the next chapter is solved.
• 17:03
  SPEAKER_00: That attitude gives us room to keep learning while holding on to standards that matter outside of demonstration.
• 17:10
  SPEAKER_00: And there is plenty to be excited about.
• 17:12
  SPEAKER_00: So imagine a small organization able to explore ideas that once stayed in a notebook because it lacked time or technical help.
• 17:20
  SPEAKER_00: Imagine somebody learning a difficult subject with explanations they can keep questioning.
• 17:25
  SPEAKER_00: Imagine researchers testing promising possibilities with better tools while doing the hard work of validation.
• 17:31
  SPEAKER_00: Those possibilities are worth pursuing.
• 17:33
  SPEAKER_00: The human payoff might be an invention, a stronger business, a new skill, or simply more time for someone who has been stretched too thin.
• 17:42
  SPEAKER_00: Bring this down to your own week.
• 17:44
  SPEAKER_00: Pick one low risk task.
• 17:46
  SPEAKER_00: Use information you're allowed to share and compare the result with your usual approach.
• 17:51
  SPEAKER_00: Check the output and count the review time.
• 17:54
  SPEAKER_00: That's a useful beginning for a worker, a business owner, or somebody who's simply curious.
• 17:58
  SPEAKER_00: If you'd like to learn to test these ideas with me, Santaclaritaai.com is a place to start.
• 18:04
  SPEAKER_00: The point is to develop enough firsthand understanding that can ask better questions about what comes next.
• 18:11
  SPEAKER_00: The transition moonshot starts with a commitment to take people as seriously as capability.
• 18:18
  SPEAKER_00: We can produce astonishing technology and put serious effort into paid learning, meaningful choice, accountable deployment, and shared benefits.
• 18:26
  SPEAKER_00: We can prepare for more powerful systems without pretending we know the arrival date.
• 18:31
  SPEAKER_00: So return to that room where the most capable intelligence might be something we built.
• 18:36
  SPEAKER_00: What kind of life do we want that capability to make possible?
• 18:40
  SPEAKER_00: Who gets to help decide?
• 18:42
  SPEAKER_00: Those are questions we can start answering together right now.
• 18:46
  SPEAKER_00: I'm Connor with Honor.
• 18:47
  SPEAKER_00: Thanks for watching and be safe.