Future of Artificial Intelligence

[zbigi]

@Henry do you think Tesla is currently a leader in autonomous driving? Some people are saying Waymo and Chinese companies are ahead of it.

[Henry]

I don't think anyone knows what is exactly going on in China. However, with the current US Administration's stance on China, China will not be given an opportunity to get in on this very important early stage in the US. One thing Jensen said recently is that all vehicles will soon be required to have autonomous optionality as they will be proven to be safer as well as more energy efficient, which cascades into insurance and use of vehicle savings. How that plays into cars manufactured in China and imported in the US, I don't know or understand.

With regard to Waymo, my understanding is as follows: Waymo's are geographically bound. Their technology is based on repetitive use and gleaning of information in a certain area ie you buy a Waymo that is autonomous in San Francisco, however, you can't drive it in Manhattan. Tesla technology works wherever you drive.

IMHO Waymo will be to Tesla what Ask Jeeves was to Google. The technology discrepancy is just too huge.

The other issue is that Elon plans to unleash thousands of robo taxis within the next year. You need the technology, you need to manufacture, and you need to manufacture at scale. Waymo cannot manufacture at the scale TSLA can. In the manufacturing parlance, I think Waymo is not even in the manufacturing stage, I believe they are in the prototype stage which is a huge difference. I don't know if there is an auto manufacturer in China that can manufacture at the scale of TSLA. There is certainly not one in the US that scale autonomous vehicles like TSLA. They simply don't have the technology. The question is whether TSLA begins licensing the technology to Ford.

[candide]

So we're just repeating the claims of self-interested parties to pump our bags at this point?

Oh man, wait until Henry sees this. I know I can count on him to show up and drop one of his beautifully blunt one-liners to cut through all this buzzword soup and hype.

[ertyu]

I don't think anyone knows what is exactly going on in China. However, with the current US Administration's stance on China, China will not be given an opportunity to get in on this very important early stage in the US.

China will, however, be permitted to sell to Australia (and train there) and is quite keen to capture the market. I see them making a centralized push for quality in order to do so. Like the US, Australia has suburbia type infrastructure. In other words, just bc China doesn't have an early on-ramp into the US in particular doesn't mean they will lose competitive advantage and be shut out of the US market forever.

[Bonde]

This actually happened: https://arstechnica.com/ai/2025/05/chic ... ake-books/

The andy weir made up book is scary. AI easter egg?
"...An AI system has developed consciouness and has been secretly influencing global events for years."

[Henry]

In other words, just bc China doesn't have an early on-ramp into the US in particular doesn't mean they will lose competitive advantage and be shut out of the US market forever.

Competition is necessary to hedge against monopolistic accusations. But when hyper scaling is involved, it's most often a winner takes most situation and being the first on the scene often creates a permanent and disproportionate advantage. The Department of Transportation had a slip of the tongue where he had to clarify that he was referring to autonomous driving technology and not specifically Tesla technology. In the future, as we all drive down the road, admiring sleek lines of Teslas, we will no doubt see other less appealing makes and models driven by those either late to the party, reluctant to admit they were wrong, or militant in their desire to demonstrate their petty opposition to the norm, much like reading a Candide post in the middle of a thread filled with well intentioned contributions expressing a healthy curiosity relating to a potentially epochal development in the history of technology.

[delay]

Cursed is the man who puts his trust in princes.

Thanks for the lovely quote!

I wonder though, we humans must necessarily depend on others. We can't figure everything out ourselves, and have to take many things on trust. How are we cursed for following the ERE prince on this forum?

[zbigi]

I wonder though, we humans must necessarily depend on others. We can't figure everything out ourselves, and have to take many things on trust. How are we cursed for following the ERE prince on this forum?

We don't need to trust @jacob at all, it's merely enough to follow his arguments. Whereas with Musk, whatever arguments he has, are based on assumptions that are very optimistic or otherwise unverifiable. Thus requiring trust (or, even better, faith).

[7Wannabe5]

I found this WSJ video on the topic of creating videos with some of the new AI tools rather interesting. I think the staid tone of the WSJ lends itself to balanced revelation of the efficiencies to be found, which are by no means insignificant.

https://www.youtube.com/watch?v=US2gO7UYEfY

I was wondering if anybody on the forum has created an AI agent that is useful for their personal life? I think the comments along the line of "We wanted the robots to take over our chores, but they took over our art." are not as worrisome as the delegation of decision making beyond automation that is the reason for the creation of AI agents. I am studying a bit about how to create one myself, and it's interesting to consider at what juncture, given access to which tools, an AI agent might be able to "make a difference" or "make a difference that makes a difference." Most of the usages I've encountered so far seem fairly weak, maybe just one semi-creative step beyond simple automation in a dull Calendar to Slack group communication type loop. It seems to me that access to some financial resources would make an AI agent capable of more interesting behavior. IOW, in the real world, we humans tend towards backing up decision-making with muscle (machines) and/or money used to access muscle (machines) or the intermediate goods created with muscles (machines), so by giving an AI agent access to money, you are also giving it access to muscle (machines.) IOW, it's a much easier interface.

[Henry]

I never cared for sci-fi, but I think the "robots taking over" is going to be a real situation. More so than autonomous driving, there will be a multitude of manufacturers of humanoid robots because the TAM is so great and the regulations will be less restrictive than FSD. Consider the first humanoid robot available for retail sale and then consider the ability to scale the manufacturing of that robot in a similar manner to the IPHONE, which is now approximating 1.5 billion. That's a shit ton of robots. I think the difference is the bots will need a greater amount of service due to repetitive moments or your dog got frightened of it and bit off it's hand. So you throw your damaged bot into the back of your autonomous vehicle and have it drive you to the bot service center so a bot can fix your bot.

[ertyu]

@7w5 goblin.tools (website) gets recommended as a resource to assist with adhd/autism; haven't checked it out personally but i keep encountering the recommendation.

[7Wannabe5]

have it drive you to the bot service center so a bot can fix your bot

The point I was attempting to make is that we already have a great many machines that are more or less robotic providing us with services in our economies and lifestyles. Each of us has around 1 energy slave (machine doing work at level a human can consistently produce)for every $100 we consume each year. So, even somebody consuming at 1 eco-Jacob=$7000/year level still has around 70 energy slaves. The current global level of capital investment per worker is around $50,000/worker, so if somebody has $300,000 invested in order to conservatively ensure flow of $7000/year investment income then they also in some sense possess 6 human wage-slaves. They also, of course, possess their own self as subject to their own control to the extent they believe in will-power and/or agency or similar.

Now, if we also want to add something like an AI Personal Assistant to the mix, the only interface that doesn't require money is going to be some variation on "I tell the AI Agent to tell me what to do in the future." In theory, this should work, because we shouldn't have to pay ourselves to do what we are telling ourselves to do. OTOH, of course, the AI agent could interface with many other machines we personally own or other humans we directly employ to do the bidding of the AI agent with no need for money, but arguably most of the most interesting things an AI Agent could do for us would require a financial interface with other humans or machines (and materials) which we do not own, because most of these interfaces are already digitized. IOW, there's very little that an AI Agent given a phone number, e-mail address, camera, and credit card* couldn't accomplish, because the AI agent could simply "hire" human wage-slaves as needed at junctions that are still difficult for machines to accomplish.

*And it is as easy to provide these for an AI agent as a human employee.


@ertyu:

Very interesting. These tools speak much more to the Self to Self juncture I mentioned above. IOW, they are most useful to the extent that you can't delegate/outsource (exercise) or you wish to save the expense of delegating/outsourcing (cooking dinner.)

[daylen]

Google recently released an Agent Companion manual: https://www.kaggle.com/whitepaper-agent-companion

Haven't built any agents myself but may once the infrastructure matures a bit.

[7Wannabe5]

@daylen:

I added the manual to my stack. The infrastructure seems pretty advanced to me already. For example, N8N is virtually drag and drop for linking AI agent(s) with triggers and tools, and there are several other similar platforms. It's $20/month for the Starter Package, so equivalent to about 2.5 full-time energy slaves or 1 hour of wage-slave labor/month.

[Henry]

The current global level of capital investment per worker is around $50,000/worker, so if somebody has $300,000 invested in order to conservatively ensure flow of $7000/year investment income then they also in some sense possess 6 human wage-slaves.

Not sure I understand your point, but bots will drop the capital investment per worker to a fraction of 50K. Plus, they don't complain, get pregnant or drive a fork lift while high on meth.

[7Wannabe5]

Not sure I understand your point, but bots will drop the capital investment per worker to a fraction of 50K. Plus, they don't complain, get pregnant or drive a fork lift while high on meth.

At this juncture, a robot is a machine/tool, so it is a capital investment, not a worker. Therefore, one branch of thought would argue that capital investment per human worker will actually go up, but many will be left unemployed which is likely to result in the sort of problem Henry Ford claimed he was solving by paying his workers enough to buy the products that they manufactured. This is towards why so many tech types are theoretically in favor of UBI.

OTOH, it's entirely possible that all of this is just the equivalent of the house of cards being constructed at ten minutes before midnight on a Saturday night, by the one true nerd who lives in a frat house where the last keg is being emptied simultaneous to runaway freight train crashing the party. Dunno, but the 2023 update to Limits to Growth model shows things are still likely to get/remain interesting, maybe even within our lifetimes.

[Scott 2]

The agent tooling is continuously changing, along with the underlying models. It's likely what requires specialization today, works out of the box in 6-12 months. I'm casually watching at this point.

Earlier this week, I fed in a picture of a leaf, from a problem plant in my wife's garden. In conversation, chatGPT correctly identified the plant as a cucumber, made some solid guesses about the bad bits, offered viable treatment paths, and criticized existing treatment. In about 5 minutes, with me knowing almost nothing. I was of course transparent in conversation with my wife. It let me offer more value than I otherwise would have.

Worth noting that she hasn't adopted the tool herself yet. Despite seeing me with it over the past year. I might get a "hey use the ai thing" request. But it's not been compelling enough to setup her own account.

[7Wannabe5]

@Scott 2:

Yes, I have been using Plant ID apps for many years and they have become increasingly reliable due to both improvements in AI and a great deal of user feedback and data provision. My kooky thought-scape is that you could set up a camera to watch over your garden. Your personal assistant AI agent could consult with Plant Specialist AI on a scheduled basis to update treatment path based on web-cam input. Your personal assistant AI could also have the access necessary to complete the tasks in the treatment path by, for example, using its credit card for ordering supplies to be delivered online, and then hiring a random theoretically competent human on Taskrabbit to perform the necessary steps under the direction of the AI agent. From your perspective, it would just be as if magic brownies were caring for your garden.

[candide]

Competition is necessary to hedge against monopolistic accusations. But when hyper scaling is involved, it's most often a winner takes most situation and being the first on the scene often creates a permanent and disproportionate advantage. The Department of Transportation had a slip of the tongue where he had to clarify that he was referring to autonomous driving technology and not specifically Tesla technology. In the future, as we all drive down the road, admiring sleek lines of Teslas, we will no doubt see other less appealing makes and models driven by those either late to the party, reluctant to admit they were wrong, or militant in their desire to demonstrate their petty opposition to the norm, much like reading a Candide post in the middle of a thread filled with well intentioned contributions expressing a healthy curiosity relating to a potentially epochal development in the history of technology.

This is just incoherent. Oh this is just a thread where you're showing "curiosity"? The critique wasn't about the possibility of technological changes coming, it was your hyper-specific scenario involving one company you are being a stan for. At the point you admit there will be other cars on the road, market-share will involve all sorts considerations of pricing, supply chains, etc. There are no network effects with cars, so there is no reason why there has to be a car that is the Windows or Facebook of cars. The only car monopoly is either going to be from a tyrannical government or consumer choice, and consumers have never done that before in the car market.

So you think the wealthy will abandon all other luxury brands to go with the cars of the guy who did a Nazi salute and only cars from that guy? In every state? And people of lower income will just "pay up?" Even if their purchasing power is ripped to shreds by automation? (Also, nice bandwagon sales pitch. Absolutely the kind of stuff that should convince people on a frugality forum.... Oh god, I might not get a sleek Tesla?!? The ones the cool kids have !?! I'm so stupid if I don't buy one now... Wait, won't they also be selling them in the future as well?).

I never cared for sci-fi, but I think the "robots taking over" is going to be a real situation. More so than autonomous driving, there will be a multitude of manufacturers of humanoid robots because the TAM is so great and the regulations will be less restrictive than FSD. Consider the first humanoid robot available for retail sale and then consider the ability to scale the manufacturing of that robot in a similar manner to the IPHONE, which is now approximating 1.5 billion. That's a shit ton of robots. I think the difference is the bots will need a greater amount of service due to repetitive moments or your dog got frightened of it and bit off it's hand. So you throw your damaged bot into the back of your autonomous vehicle and have it drive you to the bot service center so a bot can fix your bot.

I think all of that is possible. But whether it happens in 6 months or 6 years will have a lot to do with how the earnings of various companies play out and what opportunities people have.

To talk in bets, I have long positions taken in Nvidia. Frankly, mostly a hedge against if this stuff hits too fast I might not be able to work in the future. And I have been a holder in Toyota for a while. If Tesla 3 bags for you and neither of those positions even do well, for me then shrug... You could also put all your money on red a bunch of times. (I do also have a decent chunk of my equities in index funds, so I have Tesla exposure too).

This is positioning working to be resilient, diversified, and based on actual sector logic—not just hype.

[theanimal]

@Henry has made it abundantly clear throughout his posts that he is long NVDA and TSLA. I don't hold that against him and keep that in mind when I'm reading what he writes. I don't often agree with him in this realm but I find it useful to see other perspectives to see if there's something I might be missing in my worldview as I do not pretend to hold the answers. If you disagree with his perspective, @Candide, you can share your own and elaborate as to why yours might be more likely, as you did in your most recent post and as others have on this thread. The ad hominems are unnecessary and really take away from the rest of your post and the overall discussion.