Future of Artificial Intelligence

[Henry]

The California riots revealed an interesting tidbit on Waymo - the destruction of the five vehicles cost Waymo $750K to $1M. So at this point, a Waymo costs somewhere between $150K to $200K. That's Waymo money than a Tesla. My understanding is that the lidar systems including cameras and mapping systems cost approximately. $90-100K per vehicle. Furthermore, my understanding is that Waymo does not plan on manufacturing the vehicles, hence their recently announced partnership with Uber. The sale of Tesla autonomous vehicles or any other autonomous vehicle is a typical car sale with the hope of sell a customer a car once, sell a customer cars for life. The business model changes with Robotaxis. At the current cost per mile, $4.00 to $1.50, the TAM of robotaxis is estimated at $90Billion. The inflection point for an exponential increase in TAM comes when the cost per mile is reduced to $1 or less because at that cost per mile number it would just not make economic sense for a wide swath of the population to own a car anymore and it would create a network transition to the service model. To get to $1 or less per mile, the unit economics of the car is the primary determinate. Juxtaposing TSLA's ability to mass manufacture their own vehicle with the installation of their own software system in the manufacturing process at an estimated $30K per Robotaxi (TSLA just manufactured their 8 millionth vehicle 9 months after manufacturing their 7th million vehicle) as opposed to Waymo's $100K operating system being installed into a fleet of vehicles owned by a partner, I'm viewing Tesla as having an inordinate advantage in getting to that $1 or less cost per mile. Elon providing today's well timed mea culpa to The Big Guy before the possible June 22 Austin reveal puts that nonsense to rest and gets the focus back to the issue at hand.

The drone manufacturing issue is problematic. Elon has stated the Ukraine war is primarily being fought with drones and he does not want TSLA to become a US contract military manufacturer so not sure where that is heading.

[7Wannabe5]

There's also the possibility that the success of AI will lead to the transcendence of capitalism itself. Due to my internal compulsion to expand debate to furthest horizons, I am currently reading, "Fully Automated Luxury Communism" by Aaron Bastani. He points to a 2011 article in The Economist which posed the question, "What happens when...machines are smart enough to become workers? In other words when capital become labour?" as a sort of inflection point. Also, the fact that as Stewart Brand declared in 1984, "Information wants to be free", the decreasing towards zero cost of transmitting information, has contributed to the existence of bloated whales such as TSLA dominating the public-equity market. IOW, to some extent these behemoths are protecting their intellectual property by virtue of their very size, and this reality is also making them increasingly dependent upon state support, due to state-held monopoly on violence. IOW, I don't know whether we are still in a space where this firm vs. that firm is the appropriate focus, but that is not to imply that it is time to start investing in personal armaments and meat rabbits.

[Henry]

Elon Musk says 30% chance of everything going sideways, 70% chance of pass the hookah on a beach anyone want to go to Mars this weekend post-need to go to work let the machines run the place world. Others as well are speaking of a complete reimagining of the current order(s) political and otherwise. AI controls, similar to Cold War arms controls, will most likely arise with potential limitations on compute increase. But most importantly, TSLA is not a bloated whale. It is in fact, a nimble forest creature, twirling innocently in a gentle rain, preparing for its gracious leap to $4,500 per share. If at that time it does reveal itself to be a holy terror, so be it. I'll deal with it then. Let's just have a problem free Robotaxi rollout and put these human extinction concerns off until 2028.

[7Wannabe5]

@Henry:

I agree that future reasonably resembling the past for the next 3 years is likely. OTOH, all bets but one off by 2055. IOW, I am only willing to bet that 2055 will differ from 2025 more than 2025 differs from 1995. Hopefully, we can both make it to 90 in order to find out.

[Henry]

I think 2028 will be an inflection point, and it will be closer to 2035 then 2055, which gives us a better chance of seeing whether it will be sirens or stars in our eyes although I'm set to exit stage left in 2033 but maybe the woo woo world can do some negotiating on my behalf because I've been such a good fucking boy and stayed out of their subversive hair all these years despite that one time with the I think she may have been a hooker that looked like Marcia Brady where some sort of supernatural shit was definitely going on. The issue is not the technology or the ability to scale, its energy. The current energy model will not suffice. TSLA is AI, transportation and energy storage. The third will be the most challenging for the obvious reasons. It's the idea that it will not be until 2055 that has everyone fooled. I think most people are acclimating to the "not if" but are extrapolating the "when" too far out into the future. I might be early but I think it will be much less than 30 years.

[zbigi]

@Henry:

I agree that future reasonably resembling the past for the next 3 years is likely. OTOH, all bets but one off by 2055. IOW, I am only willing to bet that 2055 will differ from 2025 more than 2025 differs from 1995. Hopefully, we can both make it to 90 in order to find out.

Didn't people in the sixties expect that, by year 2000, we'll colonize other planets, or, at the very least, have a moon base, flying cars and other fantastic things that never materialized?

[Henry]

Didn't people in the sixties expect that, by year 2000, we'll colonize other planets, or, at the very least, have a moon base, flying cars and other fantastic things that never materialized?

Only when they were on Acid.

[7Wannabe5]

Oh, I meant 2055 for something other than capitalism as we know it due to meta-crisis inclusive of AI, not for semi-Jetsons lifestyle for top 10%. I agree that could come much sooner. Of course, the 30 million humans still engaged in nomadic herder lifestyle in Africa still haven't come to learn capitalism as we know it, so "future is already here, just not evenly distributed" will likely still hold. The interesting thing about the Jetsons and the Flintstones is what a crappy job they did in predicting social/cultural trends. Wilma is a prehistoric 1960s housewife and Jane Jetson is a futuristic 1960s housewife. When we no longer need to be intelligent, maybe our robots will compete at how pretty they can make us.

[delay]

Didn't people in the sixties expect that, by year 2000, we'll colonize other planets, or, at the very least, have a moon base, flying cars and other fantastic things that never materialized?

People in the 1960s also put a human on the moon and had him plant a flag. We somehow unlearned how to do that.

[zbigi]

People in the 1960s also put a human on the moon and had him plant a flag. We somehow unlearned how to do that.

The whole idea of putting a man on the moon was to show the world that the US, and not the USSR (who have just sent a first man in history into space) is a technological leader, and thus, long-term, it's smarter for neutral countries to allign with the US than with the Soviet block. It was a propaganda event/tech demo that fulfilled its purpose, there's not much needed to repeat it now, as today no one is excited about space any more (the key technological developments, such as ICBMs and sattelites are already mature technologies). Today's equivalent would be the US and China trying to outdo one another with impressive (if impractical) AI tech demos.

[jacob]

On that note ... a practical way of telling whether a new techmology actually changed how people live is whether it results in universally getting a new form of utility bill. E.g. water, electricity, gas, phone, internet, health care, ... maybe "amazon prime"? Otherwise, life is much the same.

A time traveler from 1995 wouldn't exactly have their mind blown or feel completely out of place in 2025. No new utility bills have been added.

Another way to tell is whether a new form of travel has been added. I'm afraid the possibilities for that have been exhausted.
Feet, horse, wagon, ship, bicycle, train, car, airplane, ... and going back to the 1960s people thought about rockets and flying cars but those modes are simply too expensive to turn into everyman transport relative to what you get out of it. I mean, flying from NYC to Paris in 7 hours (instead of 4 days by boat), amazing! Cutting those 7 hours down to 3 hours at 10x the price? No thanks!

Methinks the future of AI very much depends on how intelligent it gets. If it stays at the level of the "average failure prone human", it strictly becomes a question of how expensive that AI is (you guys know the numbers) relative to how expensive it is to hire such a human ($11/hour). Nothing revolutionary is going to change much unless those two numbers differ by an order of magnitude.

[zbigi]

Methinks the future of AI very much depends on how intelligent it gets. If it stays at the level of the "average failure prone human", it strictly becomes a question of how expensive that AI is (you guys know the numbers) relative to how expensive it is to hire such a human ($11/hour). Nothing revolutionary is going to change much unless those two numbers differ by an order of magnitude.

Yep, and, for many things already, it's already a matter of cost and not technical feasibility. E.g. iPhones are assembled by hand not because we can't make robots that do that, but because (Asian) people are way cheaper. Valve had full robotic assembly of their Steam Controller 10 years ago already (https://www.youtube.com/watch?v=uCgnWqoP4MM), but they're a private company, so they respond to the whims of the owner who thought it would be cool and that some profit can be sacrificed for that.

[chenda]

I wonder if everywhere having first world labour costs will be a prerequisite for widespread adopted of near total automation?

[Henry]

A time traveler from 1995 wouldn't exactly have their mind blown or feel completely out of place in 2025. No new utility bills have been added.

I think stepping into the 2025 IPHONE world would not only make a 1995er feel completely out of place but would in actuality make them completely out of place. Let's take a selfie, Venmo me, text me, did you get the Zoom link, that mailbox hasn't been here since 2005, that dirtbag in High School is now a crypto millionaire, my Amazon shipment for a single AAA battery just arrived, I don't need to stop for gas its a fucking a Cybertruck, let me ask SIRI, I never met my best friend, I spent all weekend binge watching classic episodes of Girls Gone Wild, yeah she is hot but I think she once had a penis, you can buy pot in the store where you used to make your travel arrangements, I used to enjoy this social media community filled with really smart people who wanted to get out of working until this complete asshole came along and wouldn't stop posting crass screeds about how self driving cars, traveling to Mars and robots wiping my grandmother's assholes will be here by Thanksgiving. At least to me, that seems like a lot to take in the first day.

[7Wannabe5]

A time traveler from 1995 wouldn't exactly have their mind blown or feel completely out of place in 2025. No new utility bills have been added.

It might depend a good deal on where they took off and landed. For example, in Africa, the first utility bill by the majority is cell phone bill, and the second bill acquired is monthly rent-to-own on solar-panels bundled with radio, lightbulbs, and possibly television. The acquisition of cell phone bill also provides access to bank account and similar amenities. As of 2019, this bundle was being sold in Kenya for $350, $35 down and $15/month bill. Meanwhile, water may still be simply drawn from communal well. IOW, the old school 20th century grid, order of bill received, is being entirely leap-frogged. Adoption of smart phones in Nigeria is currently as high as in the U.S. and the much faster growing population is much younger.

Nothing revolutionary is going to change much unless those two numbers differ by an order of magnitude.

Five independent AI agent workflows performing 2500 total workflow executions with unlimited steps per workflow currently costs $20/month. So, within the limited scope of activities an AI agent can currently perform, if average workflow execution would require 10 minutes for human to perform, the difference would already be $11/hour for human vs. $.05/hour for AI agent. However, this does not necessarily fully account for the expense of training the humans on either/all sides of a human/AI interface. For example, consider all the redundancies the average Western realm human now engages with in terms of the simple task of telling time. Why do you even need to know what time it is if Alexa/Siri/Much-More-Advanced-AI-Agent knows the time? Similarly, one of the largest expenses involved with the use of industrial robots is ensuring the safety of the human workers with whom they interact, as is also obviously the case with self-driving cars and the general public.

[jacob]

@Henry - All those concepts were available in 1995. There has been no/few additions in terms of WHAT and HOW you can do in 2025 that you couldn't do in 1995. All that's changed is WHERE with smartphones now dominating but not replacing the already universal "home computer" of the 1990s.

For example, depending on how you count, video calling has existed somewhere between 50 and 100 years already. It only became popular during COVID not because of technological innovation but because demand was suddenly there. Most calls between humans remain voice only. If anything, calling is being replaced by texting amongst the younger generations. The SMS system was rolled out in 1993.

[Henry]

A reductionist Ecclesiastical "nothing new under the sun" argument is always true as we got the constant of human nature thing going on, but not imputing a move from conceptualization to broad adaptation as meaningless unless you get a PSE&G bill doesn't seem to be the determinate as to whether epochal change has occurred. Books were available before the printing press, people could read before the printing press, the idea of a printing press was available before the printing press and the human potential to move machine levers was available before the printing press. I don't think anyone got any new utility bills at Gutenberg. But when it came together, a new world order was established.

[jacob]

@Henry -

On that note ... a practical way of telling whether a new techmology actually changed how people live is whether it results in universally getting a new form of utility bill. E.g. water, electricity, gas, phone, internet, health care, ... maybe "amazon prime"? Otherwise, life is much the same.

So, the rhetorical question is whether moving "internet services" from the home computer to the smartphone has made a material difference. The best I can come up with is that many humans are no longer capable of making plans in advance and have lost the ability to follow directions. Instead, people coordinate on the fly with texting and just ask google to direct them to their destination. Otherwise ...

And likewise, will e.g. "average IQ AI" do anything beyond replacing office workers with half an associate degree with an "IT system" or replace $40,000/year taxi and bus drivers with $40,000/year robo-drivers? Of course if I was a bus driver or investing in AI-bus systems, it would make a difference for me. As a passenger, I don't see much difference.

Initially, there was a naive presumption that "the internet" was going to funnel knowledge and information into every home. Much like TV and radio before that. Instead, the internet has mainly served to translate existing services onto itself. For example, streaming movies over the internet instead of DVRing them from cable. If you don't look under the hood, there's practically no difference between what you could do then and what you can do now.

Indoor plumbing on the other hand, now that is a remarkable invention.

[7Wannabe5]

I don't think anyone got any new utility bills at Gutenberg.

Well, many affluent humans did pay monthly fees to be members of lending libraries. Free libraries were actually one of the more revolutionary and remarkable American institutions. So much so, that when my "ex" immigrated from Iran in the late 70s, they still seemed wonderful to somebody who considered himself privileged in his youth due to ability to afford textbooks not available to the same age son of his Nanny beyond his "noblesse oblige" in providing them as hand-me-downs. "Information wants to be free", but it may take some time to achieve it. Of course, now it is the case that anybody with a phone and an intermittent internet connection has access to almost any book in the public domain and plethora of courses offered by top universities for around $40/month.

What's the market for self-driving vehicles for affluent old Americans vs. the market for solar-powered fat tire scooters for young African kids? Why keep building or innovating on top of an out-of-date, too expensive infrastructure?

[Henry]

Outdoor plumbing did not eliminate human shitting one turd. AI has and will, to a degree yet to be determined, eliminate human labor. I will go with the latter as being more impactful in the overall arc of human history.