Latest / eMotors: Electric Revolution / The Future Prohibition of Human-Driven Cars
Transcript
- 0:09OK, imagine this for a second. A future where getting behind
- 0:14the wheel of your car may be that classic you've lovingly
- 0:16restored. It's actually, well, illegal.
- 0:20Not just frowned upon, but genuinely against the law.
- 0:23Why? Because statistically, you, the
- 0:26human driver, are just too dangerous.
- 0:29Yeah, it's a really provocative idea, isn't it?
- 0:31And it's it's kind of the core theme in the research you
- 0:34shared. We're seeing this shift in the
- 0:36conversation about AI. It's moving away from just, oh,
- 0:40robots might take jobs because they're cheaper towards
- 0:43something more fundamental, like maybe humans are becoming a
- 0:46safety liability. Exactly.
- 0:48And that's what we're digging into today.
- 0:49This deep dive is all about that emerging idea, backed by, you
- 0:53know, research and what experts are saying that AI should take
- 0:56over tasks not just for speed or efficiency, but primarily
- 0:59because human error, our mistakes, our fatigue, it's all
- 1:03becoming an unacceptable risk. So our mission today is
- 1:06basically to unpack that. We want to look at the the main
- 1:09reasons being put forward for this big push towards AI
- 1:11replacement. Yeah, the foundations.
- 1:13Right. And then really dive into the
- 1:15data, especially around autonomous driving versus US
- 1:18humans. It's pretty stunning stuff,
- 1:20definitely. And finally tackle the, let's be
- 1:22honest, the huge social and legal bumps in the road this
- 1:26whole transition creates. It's like mapping out a future
- 1:29where maybe human flaws get legislated out of existence.
- 1:33OK, let's start there then, with the the overall idea of
- 1:37replacement. Some experts are arguing pretty
- 1:40strongly that replacing human work with machines isn't just
- 1:43inevitable, it's beneficial, and we should actually speed it up.
- 1:46Yeah, that perspective really frames the debate.
- 1:49The thinking breaks down into about four key areas, 4 pillars,
- 1:53you could say, OK, the first one is efficiency, obviously, but
- 1:56it's not just that machines don't need sleep, right?
- 1:58We all kind of get that. Sure, The real kicker is the
- 2:00consistency. Perfect output, 204, seven.
- 2:03No bad days, no off moods, no biological limits messing things
- 2:08up, Just pure predictable performance.
- 2:10That makes the historical comparison some people use quite
- 2:13powerful, like the shift from using, say, oxen for farming to
- 2:16using tractors. Exactly.
- 2:18It wasn't just faster plowing, it was predictable output you
- 2:22could rely on day in day out. And that predictability fuels
- 2:26the second pillar, speed. Right.
- 2:28Speed don't have our biological bottlenecks.
- 2:31They process information, move things, react at speeds that,
- 2:35well, our bodies just can't handle.
- 2:36We're talking microsecond reactions in really complex
- 2:39situations. OK.
- 2:40So efficiency, speed. The third one is economy.
- 2:43And this is where it gets pretty intense, I think.
- 2:46Well, yeah, because the initial cost, the investment in this
- 2:49high level AI and automation, it's huge, no.
- 2:53Question massive. But the argument is over the
- 2:56long run, they become way, way cheaper.
- 2:58Productivity goes up, maintenance becomes more
- 3:00predictable. So companies that take the
- 3:03plunge, you get this enormous competitive edge.
- 3:06It's not just saving a bit on salaries, it's potentially
- 3:08freeing up huge amounts of capital, maybe reinvesting that
- 3:12former salary fund into just massive growth or R&D.
- 3:15It could really change the game for market dominance.
- 3:18And the final pillar ties right back to our main theme today,
- 3:21safety. Safety.
- 3:23And this is interesting because the meaning has kind of evolved.
- 3:26It used to be about protecting humans from dangerous jobs, like
- 3:29getting us out of minds or off dangerous factory lines.
- 3:33Now, increasingly, it's about protecting the system from us,
- 3:36from the human element. Machines don't spread diseases
- 3:40between animals and humans, for instance.
- 3:41And crucially, they don't make those tired, distracted or
- 3:45emotional mistakes that can cause, you know, huge failures.
- 3:49And to show just how far AI has come beyond just doing simple
- 3:52tasks, the sources point to some really striking examples, like
- 3:56this AI transcription tool Whisper.
- 3:59Oh, whispers. A fascinating case he learned
- 4:01from something like 680,000 hours of audio.
- 4:04Just incredible amounts across different languages, accents,
- 4:07really noisy backgrounds, think bars, bad phone lines, you name
- 4:11it. And this take away is it's not
- 4:13just good, it's better than human transcribers, even in
- 4:16those really tough conditions. Better than a professional
- 4:18focusing hard in a noisy room. Exactly.
- 4:21That's the leap. And then you've got things like
- 4:23this other AI, Claude, which shows the cognitive jump.
- 4:26It can apparently read and process like an entire books
- 4:30worth of text in maybe 30 seconds 30.
- 4:32Seconds, yeah. Something like that.
- 4:34But it's not just the speed, it's that it can synthesize it
- 4:36instantly, connect ideas from thousands of pages, spot
- 4:40patterns, generate reports, stuff that honestly, no human
- 4:44team could do that quickly. It's described as just far
- 4:48beyond any human capacity. We're really talking about
- 4:51exceeding basic human cognitive limits here.
- 4:54OK, so let's really unpack that safety angle now, because the
- 4:57most, well, the most life and death application seems to be on
- 5:00our roads. The materials we looked at
- 5:01basically suggest human drivers are becoming intolerable, too
- 5:05dangerous. Yeah, that's where the numbers
- 5:07get really difficult to ignore, especially for regulators.
- 5:09You'd imagine, take the data from, well, from major
- 5:13autonomous vehicle developers. Their vehicles show an accident
- 5:16rate of around .59, let's say .6 accidents per one and a half
- 5:20million kilometers driven. OK, .6 per 1.5 million
- 5:24kilometers, How does that stack up against US?
- 5:27Yeah, against humans, it's. Not great for us.
- 5:29The human average is about 3 accidents per that same
- 5:32distance. 3 compared to .6. Yeah.
- 5:35So roughly A5 fold difference in risk.
- 5:37Yeah. And when you think about the
- 5:38hundreds of thousands killed and millions injured globally each
- 5:41year because of human driving errors, the argument shifts
- 5:45pretty fast. It goes from can AI drive safely
- 5:48to Can we morally afford not to let AI drive?
- 5:51That is a huge difference. But you know, got asked, are we
- 5:55sure it's fair comparison? Are those AI cars driving in the
- 5:57same messy rush hour, bad weather conditions we all deal
- 6:01with every day? Or is it mostly, you know,
- 6:03perfect sunny day testing? That's a really fairpoint.
- 6:05And yeah, early testing definitely favored optimal
- 6:07conditions, but the research indicates the improvement is
- 6:10incredibly fast. Some analysis predict autonomous
- 6:13cars will meet or even beat human safety overall.
- 6:16Like, really soon. Maybe even this year, 2024.
- 6:19That means they're increasingly operating in diverse real world
- 6:22settings. And they rely heavily on this
- 6:24metric, disengagement rates improving, right?
- 6:27What exactly is that for people who might not know?
- 6:29Right. So a disengagement is basically
- 6:31when the human safety driver who's still in the car for now
- 6:35has to take control because the AI gets confused or fails or
- 6:39encounters something it just can't handle.
- 6:41OK, so if those rates are going down year after year, it means
- 6:45the AI is needing less human help.
- 6:48It's getting better at handling the tricky, unpredictable stuff
- 6:51on its own. And if these trends keep going,
- 6:54some forecasts are projecting a pretty radical future for
- 6:58driving by, say, 2050. Yeah, the prediction is
- 7:01startling. By mid century, it suggested
- 7:03that driving your own car manually on public roads could
- 7:06actually be outlawed. It might become like, you know,
- 7:08horse riding, something you do for sport on a track or private
- 7:11land, but not for daily transport.
- 7:13Wow, just imagining telling my kids that driving used to be
- 7:16normal. It's a weird thought, isn't it?
- 7:19That something so central to our lives could become seen as this
- 7:21dangerous, outdated hobby. So why?
- 7:24What makes us so bad compared to the machines in this context?
- 7:27Well, it comes down to consistency and capacity.
- 7:29Again, the A VS the autonomous vehicles, they have sensors all
- 7:34around. They're connected, processing
- 7:36way more data about their surroundings that our brains
- 7:39ever could. 360 awareness constantly.
- 7:42Constantly and crucially, they don't get tired, they don't
- 7:45text, they don't get distracted by the radio or the kids in the
- 7:48back. OK, fair, but the biggest
- 7:51problem seems to be us just being unpredictable.
- 7:54That's a huge part of it. Think about how often human
- 7:56drivers bend OR break the rules. Speeding.
- 8:00Following too close, not signaling, weaving.
- 8:03Guilty as charged sometimes. We all are right, but that makes
- 8:06us fundamentally unstable elements in a system that, for
- 8:09maximum safety, needs rigid predictability and precision.
- 8:13Machines follow the rules, always.
- 8:15And regulators are starting to grapple with this.
- 8:17We saw that example with a major EV company losing a top safety
- 8:21rating after changing its sensor setup.
- 8:23Exactly. Even if the company argued its
- 8:25cameras only system was better based on its data, the
- 8:28regulator's push back. They want redundancy, multiple
- 8:31types of sensors, radar, lidar, cameras working together.
- 8:34It all comes back to minimizing risk, demanding layers of
- 8:38assurance, safety above all. So restrictions on driving
- 8:41aren't new. We have speed limits, seat belt
- 8:43laws, phone bands, maybe autonomous only zones aren't
- 8:46such a crazy leap. Probably not, no.
- 8:49You look at cities like Barcelona implementing super
- 8:51blocks, basically large areas where car traffic is heavily
- 8:55restricted or banned altogether for pedestrian safety and air
- 8:57quality. It's not hard to imagine
- 9:00expanding that idea. OK, this downtown core
- 9:02autonomous vehicles only for safety, for efficiency.
- 9:06It becomes a regulatory tool to manage risk.
- 9:08OK, so the data seems pretty compelling, overwhelmingly
- 9:11pointing towards a VS being safer eventually, if not
- 9:14already. But let's get it to the messy
- 9:16reality. This transition isn't going to
- 9:18be smooth, right? There are big hurdles.
- 9:20Absolutely necessary context. Yeah.
- 9:22Yeah, the tech isn't flawless, and the research acknowledges
- 9:25that. One study comparing thousands of
- 9:28accidents found that yes, a VS are safer overall, but they
- 9:32actually had more accidents in certain specific situations.
- 9:35Like what? Things like making turns
- 9:37apparently and particularly during those low light
- 9:39conditions, dawn and dusk. So the system still have
- 9:44vulnerabilities, edge cases they struggle with.
- 9:46Interesting, but the biggest barrier might not even be the
- 9:49tech itself. It sounds like it's more about
- 9:50money and lawyers. Liability.
- 9:52It's. A huge one.
- 9:53Who pays when an AV does crash? The San Francisco incident in
- 9:562023 is a complex example. An autonomous vehicle got
- 10:00involved with a pedestrian who'd already been hit by a human
- 10:03driver, who then fled. Right, I remember that.
- 10:05The AV then apparently dragged the person.
- 10:07You've got human error, alleged AV error, infrastructure issues.
- 10:11It's incredibly messy. The potential liability claims
- 10:14were estimated in the millions, like $812,000,000.
- 10:17So how do insurance companies even begin to calculate risk for
- 10:20that? How do you set up legal
- 10:22frameworks to handle litigation when the driver is an algorithm
- 10:26owned by a massive corporation? It's a legal and economic
- 10:29minefield. Which brings us squarely to the
- 10:32social friction, the jobs we talked about efficiency, but the
- 10:36flip side is displacement. Millions of people drive for a
- 10:39living. Truckers.
- 10:40Taxi drivers, delivery folks. And that's probably the biggest,
- 10:43most immediate social challenge. The experts pushing for
- 10:47acceleration are clear on the tech benefits, but the plan for
- 10:50the human fallout, it's less clear.
- 10:53Understatement. Yeah, If this adoption happens
- 10:55quickly, the job losses could be really sudden and concentrated
- 10:59in specific sectors. And unlike past tech shifts
- 11:02where you know new jobs were created need new skills, some
- 11:05worry that when AI replaces cognitive work, those new jobs
- 11:09might not materialize in the same way.
- 11:10So the fundamental question of income redistribution or what
- 11:13happens with all that safe salary money, it's just hanging
- 11:16there. Pretty much.
- 11:18It's this massive administrative and social problem that frankly,
- 11:23policy makers seem way behind on figuring out, especially
- 11:26implementing solutions in complex societies.
- 11:29And beyond jobs, there's the systemic risk, right?
- 11:32Putting all our mobility eggs in one digital basket.
- 11:35That's the other big concern highlighted, the risk of cyber
- 11:38attacks or just unforeseen glitches in these incredibly
- 11:40complex interconnected systems. It's maybe more serious than
- 11:44often discussed. Single point of failure for an
- 11:47entire city's transport. Imagine a coordinated attack or
- 11:51a major software bug taking down the entire autonomous transport
- 11:54network of a major city. It could be paralyzing in a way
- 11:57we've never seen before, far worse than a strike or a
- 12:00snowstorm. Which is partly why you see some
- 12:02governments focusing more on connected vehicles first.
- 12:05Cars that talk to each other in infrastructure maybe assist the
- 12:08human driver rather than jumping straight to full level 5
- 12:10autonomy everywhere. It's seen as a potentially safer
- 12:13interim step. So that really sums up the
- 12:15attention, doesn't it? On one hand, we have this
- 12:18powerful data, these compelling arguments that AI is just
- 12:22better, more efficient, dramatically safer, or
- 12:25potentially saving millions of lives if we let it take over
- 12:28driving and other risky tasks. Yeah, the potential upside is
- 12:31enormous. But on the other hand, actually
- 12:33doing it forces us to confront these huge simultaneous
- 12:36upheavals, economic, social, legal, systemic risks.
- 12:40And honestly, there doesn't seem to be a clear road map for
- 12:43navigating that chaos. Which leads to maybe one final,
- 12:47slightly unsettling thought. If efficiency, consistency, and
- 12:50safety become the ultimate goals, the prime metrics we use
- 12:53to judge success, and if AI demonstrably excels at all three
- 12:57far better than humans, will we eventually decide that making
- 13:01really big, important societal decisions about laws, resources,
- 13:04policy is also just too risky to leave entirely in the hands of
- 13:08flawed, biased, inconsistent human leaders?
- 13:10Wow, that's quite the thought to leave our listeners with taking
- 13:13the human out of the loot at the highest levels.
- 13:15OK, well thank you for helping unpack all of this.
- 13:17It's fascinating and honestly a bit scary too.
- 13:19It is lots to think about. Absolutely.
- 13:22And thank you, our listeners, for joining us on this deep
- 13:24dive. Your curiosity is what fuels
- 13:27these discussions. If you found this valuable,
- 13:29please do subscribe to the channel so you don't miss our
- 13:31future explorations into where technology is taking us.
- 13:35We'll see you next time.