Artificial Intelligence

[Brian Peacock]

I trust the experts that these ChatGPT things aren’t all that intelligent. I just imagine there are probably a lot of smaller problems ai will solve, and at least some of them will be significantly profitable and transformative.

On the issue of real intelligence I still haven’t read anything more interesting on the subject (admittedly I don’t spend a lot of time on it) than Dennett’s essay on the frame problem: https://folk.idi.ntnu.no/gamback/teachi ... nett84.pdf Someone, or many someones have claimed to have solved it…

Machine Learning Systems are doing some amazing things, in some specialised or specific contexts. They're also doing some dreadful things. Of course, those things were amazing or dreadful before researchers started applying MLS to them.

[Svartalf]

anyway, AI is a fricking bubble... there's a Nobel prize out there wailing that the tech we use (I'm not sure whether he means the chips or the Large Language Model based programming) is totally the wrong one.
To boot, the CEO of IBM has demonstrated that the economic model just doesn't work, the investments have been so massive there's just no way to make the thing profitable, ever.

there's also been many people dissing the LLM type of 'intelligence', because it will never be really intelligent, even if it can fool laypeople.

[macdoc]

Being profitable is a side issue. The societal impact will be verra large

[pErvinalia]

It may yet be profitable given the immense boost in productivity it will drive.

[pErvinalia]

You'll all change your tune when AI solves global warming, explains abiogenesis, and comes up with a grand unified theory.

[aufbahrung]

It gives next weeks lottery ticket numbers now, although there is a two week wait to access them

[Brian Peacock]

You'll all change your tune when AI solves global warming, explains abiogenesis, and comes up with a grand unified theory.

But will it ever make the perfect Baked Alaska?

[Brian Peacock]

Being profitable is a side issue. The societal impact will be verra large

The development and implementation is wholly dependent on returns to investors - unless you're advocating that govts should take AI into public ownership?

[pErvinalia]

Mac's a communist at heart?

[Sean Hayden]

LOS ANGELES (AP) — When an AI-generated country song called “Walk My Walk” hit No. 1 on Billboard’s country digital song sales chart this month, it was credited to a fictional artist named Breaking Rust — a white, digitally generated avatar that didn’t exist two months ago.

But the song’s vocal phrasing, melodic shape and stylistic DNA came from someone who does exist: Grammy-nominated country artist Blanco Brown, a Black music artist who has worked with Britney Spears, Childish Gambino and Rihanna.

And he had no idea.

“I didn’t even know about the song until people hit me up about it,” said Brown, whose 2019 country rap hit “The Git Up” helped usher in a new, hybrid era of country crossover. He didn’t learn about the chart-topping AI track until his phone was flooded with messages from friends.

https://apnews.com/article/walk-my-walk ... 0e2998cfe2

This is the kind of thing that might make you wonder if you’re already dead, or waking up in the matrix. It’s too weird to be real…

[JimC]

An interesting article from a commentator I admire:

Is 2025 ending with an AI bubble? We could be at a turning point in history
By Alan Kohler

As we limp, downcast, into the end of another year, it's hard to avoid the conclusion that 2025 has been a momentous one, perhaps even a historic turning-point year.

Not just for the terrible way it ended at Bondi Beach, or even the way it began, with the inauguration of Donald Trump in January, heralding a wild, unpredictable lurch in the governance of the United States and the world.

Economically, 2025 is arguably up there in importance with 2008, 1973, 1944 and 1929, not because there was a stock market crash, or an oil shock or a deal by finance ministers to create a new global financial system. None of those things happened.

What did happen, I'm going to argue, was a civilisational earthquake, and that's leaving aside the fact that coal consumption hit a record high this year, virtually sealing the prospect of a climate disaster.

As for the stock market, it is quite possible that some kind of collapse is already underway because the US market touched a valuation two weeks ago that has only been seen once before in history, and that didn't end well.

That is, the cycle-adjusted price earnings (CAPE) ratio hit 40.

The CAPE, invented by American economist Robert Schiller, divides price by the 10-year average of earnings, rather than the usual one year, past or future, and is regarded as the most reliable way to measure value over time.

The previous time the US market got to a CAPE of 40 was December 1999, near the end of both the millennium and the dot com bubble, which crashed soon after. Every other time since 1880 it got above 25, the market crashed.

If the US S&P 500 did not peak on December 11 at 6,901 — which was the first time above 6,900 — and regains that level and keeps going higher, it means investors believe this time is different.

It's always different, of course, but the question as 2025 ends is: is it different enough? And that leads us to examine the thing behind the bubble — AI.

AI is changing the world, fast

So far, AI is the fastest-growing technology in history in terms of adoption.

According to a state of AI report from Air Street Capital, paid business subscriptions for AI doubled in 2025 from 22 per cent of companies to 44 per cent. A survey by McKinsey & Co puts the total business use of AI, paid and unpaid, at 88 per cent.

ChatGPT tells me it's now getting about 2.5 billion requests (prompts) per day. Nearly half of US adults use AI to search for information; 60 per cent overall, and 74 per cent of those under 30.

Spending on data centres, in which the computing for AI is done, is expected to top $4.5 trillion and is already putting enormous pressure on supplies of electricity and water, and is helping push coal consumption to its new record.

So AI is definitely changing the world, very quickly, but investment in it is also unprecedented, and the returns required to justify it are beyond anything that seems remotely likely.

Most technologies or industries where prices go beyond what's rational do end up changing the world — African slavery (the South Sea Company, 1730), railways (1840s), electricity, aviation and radio in the 1920s and the internet (1990s).

As US investor Howard Marks wrote a fortnight ago, there are two distinct aspects to bubbles: "One in the behaviour of companies within the industry, and the other in how investors are behaving with regard to the industry."

The excitement of investors serves a purpose: they provide the capital to fund the technology for the benefit of the rest of us, which they then lose because the returns disappoint.

Will that happen with AI? Almost certainly, it's just that no-one knows when.

Enter humanoid robots

But it's complicated, and possibly extended, by the second, related technology boom of 2025 — humanoid robots, which are not yet part of the stock market bubble because the companies behind them are mostly still private.

Mechanical robots have been replacing human workers for decades, but making them look and operate like a human body with an AI brain so they can think for themselves is a whole new level of human-replacing technology.

The idea is just four years old, but they are already being shipped to customers in an unconstrained global race between at least 60 companies, mostly in the US and China, but also Europe and the UK.

As with EVs, private space travel and AI itself, it was Elon Musk who kicked it off in 2021 when Tesla announced its Optimus project. A prototype was shown in 2022, and Musk said that Optimus "has the potential to be more significant than the vehicle business".

Here's a video of Optimus putting out the garbage, vacuuming the floor and cooking. Tesla has moved into pilot production and is targeting a fleet of about 5,000 of them by the end of this year.

Unitree of China is leading in the delivery of them, having shipped thousands of its G1 humanoid, which sells for $24,000 — the cost of a (very) cheap car — while another Chinese company, Agibot, delivered its 5,000th general-purpose humanoid this month.

A private American company called Agility Robotics is actively shipping to major logistics customers like Amazon for warehouse work, and two other US companies, Apptronik and Figure AI, are shipping units for "pilot deployments" to Mercedes-Benz and BMW to use in their factories.

A month ago, in a column here, I explored the implications of wholesale human worker replacement by these things, both in terms of the tax and welfare systems and the meaning of life.

Humanoid robots represent the first credible pathway for capital to operate with minimal reliance on human labour.
I wrote: "The fundamental challenge is that our tax system is built around the assumption that labour is the primary source of income. When labour shrinks, so does the tax base, and we don't have tax structures designed for trillion-dollar robot fleets. More than that, our whole lives revolve around work.

"It provides routine, purpose, identity, and community. If millions are no longer required to work, we must find alternative sources of meaning: self-education, creativity, civic life, or things we haven't thought of yet."

Howard Marks finished the essay referred to above with a postscript discussing this same issue, saying he finds the outlook for employment "terrifying".

He agrees that the problem is not just where governments will get the money to pay unemployed people a "universal basic income", but also the fact that "people get a lot more from jobs than just a pay cheque".

"A job gives them a reason to get up in the morning, imparts structure to their day, gives them a productive role in society and self-respect, and presents them with challenges, the overcoming of which provides satisfaction. How will these things be replaced? I worry about large numbers of people receiving subsistence checks and sitting around idle all day."
Why is the 2025 bubble unique?

The technologies behind bubbles are not always unqualified benefits for society like electricity, aviation and radio were. Slavery, for which the South Sea Company had a brief monopoly, definitely wasn't, and the internet is mixed — mostly good, but we're being swamped with misinformation, and the Australian government has just banned social media for under 16s.

What makes the 2025 stock market bubble unique in history is that if it's not a bubble, and the prices do end up being supported by revenue, the impact on humanity could be far more challenging than any previous technology.

If AI and data centre businesses and robot manufacturers do sell enough product to justify the trillions they're spending to build capacity, and what investors are spending on their shares, a lot of humans will be made idle.

2026 should be the year when governments think about this seriously and prepare for it, but that doesn't look like it will happen.

So far, governments are just handing water bottles to the race contestants.

Alan Kohler is finance presenter and columnist on ABC News and he also writes for Intelligent Investor.

[pErvinalia]

I think there'll be a couple of steps. First, unemployment will go up, and conservatives will get riled up by "welfare slackers". But there'll come a point where even conservative moral outrage will be replaced by practical considerations. The question is how much will we get away with taxing companies an AI dividend to pay for a universal basic income. I expect not enough. And most people will only exist just above the poverty line.

[pErvinalia]

And that's only if AI doesn't go rogue and kill us all in the meantime.

[pErvinalia]

https://www.theguardian.com/commentisfr ... uv2rbACJAw

[Tero]

The AI stuff is probably costly. We just get to play with free stuff now, but eventually the real users will be governments and corporations. The speed that they use to collect info and "think" for them.

We already had the data search part in chemistry in 2000. I never needed physical journals anymore, until I got to the paper I need. If it was an American journal, I printed the paper from it at my desk.This system for ACS (also owns it) members is 180 a year in a light version. An institution such as a chemistry department pays 6000 a year. Covers all staff and grad students.