The former head of the MIT Artificial Intelligence lab, said, in an open courseware lecture, linked here at between 14 and 16 minutes, said they don't know why it works and that they were astounded by the results produced by Geoffrey Hinton who introduced the first vastly larger neural network and ushered in the modern era of AI.

Geoffrey Hinton, associated with University of Toronto (he was Ilya Sutskever's phd advisor) has been warning about the risks of AI for some time.

The neural network roughly simulates how the brain processes information. It is the general design that nature came up with and now man is harnessing (the first commercial neural networks were used in the 90s for check scanning OCR).

They know that scaling up can eventually create a superior information processing system to the human brain. They don't know for sure. But it's a risk. THAT'S what they're concerned with primarily. If you listen to Hinton, social upheaval is also a risk.

Yes, big companies DO use all their tools, like purchasing politicians and using regulation, to squelch competition and get an advantage. That anti-competitive behavior should not obscure the risk of AI. No one knows why the LLM's can produce the results that they do. They think scaling up will make it more powerful than a human brain - but maybe it won't. I saw a study which said LLM digital computing consumes 900 million times more power than biological computing. Maybe that will limit it - until they make it more optimized.

The LLM is a digital system. It has no nervous system, no feelings. It processes information. It roughly approximates the way nature processes information with brains. They need to get a handle on what the risks of using a super-powerful information processing system like this will do.

If you want to understand the real risks, it pays to be skeptical of these companies. The best bet is to listen to top academics who are not commercially associated. Hinton is the top among them. I don't agree with all his theories - he misses, it seems to me, that the system has no nervous system. But when he talks about its information processing capabilities and what that could mean for humanity once that system becomes more powerful than humans is worth listening too.

People need to be skeptical, but not too clever by half, when thinking about the risks of AI.

People who live in condos, townhouses, apartments, single family homes without driveways - those people don't have a place to charge their vehicles at home. The optimal use case for an EV is a house with a garage and a charger in the garage. That group is wealthier, but more limited in number. What vehicle is that group going to choose? Single family homeowners with garages are probably going to go for higher-end electric vehicles. Hondas are a largely reliable economy to mid-range cost vehicle, for which that first group I listed above is the customer base, it seems to me. Honda might just be catering to their target audience.

The Inuit likely have different dietary requirements than a Mediterranean tradesman which is likely different than a south Asian farmer.

I doubt Inuit could succeed on a vegetarian diet like a south Asian farmer could.
I doubt a south Asian farmer could succeed on a carnivore diet like an Inuit could.

I don't see much discussion of this possibility when discussing diet. Nutrition discussions seem to treat all people as identical and they are most assuredly not.

In the rental market, there are organizations that manage landlord IT needs. They discovered they have access to material nonpublic information on pricing and demand, and so could "suggest" rents based on that, acting as a "super landlord". A key player is RealPage which recently settled with the DOJ about that collusion on favorable terms.

In the car dealer market, they have such organizations too, so having a "super dealer" that helps dealers price vehicles is almost certainly colluding as well.

And why is lying so... tolerated in the US?

I think the answer comes from who runs for officel. It requires someone extroverted, good at speaking, able to fundraise, and who has free time to engage in campaigning. People that fit that bill are salespeople, lawyers and career politicians. So they are over-represented. And they bring with them a certain flexible relationship with the truth. State-level politicians not infrequently graduate to the federal/national level.

I found this link about the demographics of Congress, for each Congress up to 2019 (p. 29): https://www.brookings.edu/wp-c...

Because of the Citizens United decision, which said any legal construct can contribute unlimited amounts to politicians (indirectly via PACs, super PACs, etc), fundraising has become a key element. Those who have access to professional fundraising networks (which is a thing - find an established state legislator and google "fundraiser for [state legislator]" and you can get a glimpse of these networks. I don't know how it is in Europe, but because legal constructs (unions, businesses, etc) can contribute instead of limiting it to physical persons, and not forcing political advertising to be identifiable to individuals, money in politics has become the predominant factor.

Very wealthy people in developed nations will often have large families - see Musk, Trump, Zuckerberg. When money is not a concern, many people will choose to have large families.

I've noticed the possibility of a bimodal distribution curve (i.e. like a pair of boobs, Tetons, etc), where lower income has more kids, middle income has fewer kids, and wealthy have more kids (y axis = number of kids, x axis is income).

From my anecdotal observations, I've seen people at the low end of income have a lot of kids because they're largely low-investment parents. Middle and high income are high-investment parents. Middle can't do proper high-investment (have a parent stay at home, live in a nice area so kids can go to good schools, etc). Wealthy can do all those things.

In absolute numbers, there are less high-investment capable wealthy than low-investment poor so the absolute number of children would skew towards the poor.

The "hollowing out of the middle class" could be due to a lot of factors. One significant one might be that middle tier skills yielding middle class incomes are becoming obviated due to technology.

Maybe also a larger number of middle tier ability people.

Also, DOGE savaging the federal government and contractors didn't help, because I think that's where a lot of middle tier people were or sought to be. That job and those sectors are partially jobs programs for that tier. I'm a little skeptical of UBI and more in favor of jobs programs because I think people need activity and missions. Indolence has its own dysfunctions. A "bloated civil service sector" is the bogeyman of a lot of center-right economics, but... with advancing technology, what are these people going to do? You force a middle tier type into retail customer service and he's not going to reproduce. As far as trades go, that's skilled physical labor and again, it takes a particular sort to want to do skilled physical labor, and the office clerk who is less physically oriented and more intellectually oriented is not going to want to do that, and if he has to, he again is not going to reproduce.

We need to see some real non-agenda-driven data on the issue, but these are some possibilities.

What exactly was accomplished here?

The big story that Western media is missing here is Iran's proxies, and their effect on Israeli security.

Iran's proxies are effective. I watched the Joe Kent interview with Shawn Ryan, and they admitted that Iran's proxies had killed a lot of Americans in Iraq. What is well known in Israel I'm sure, but what they've been very poor at conveying in the Western media is that they were being hit from the south by the Houthis (Shia), from the west by Hamas (Sunni) and from the north by Hezbollah (Shia). All these are Iranian proxies (Hamas is Sunni and Iran is Shia but they still work together, receiving support from Iran). Think about why Iran again closed the Strait today due to Hezbollah being hit in Lebanon.

What the Kent interview made clear was this was a war about Israeli security (also with a strong dash of hubris from the Venezuela success). Killing Iran's leadership was an Israeli statement. The Strait of Hormuz and refugee flows are not an Israeli concern, as they perceive they are facing existential threats.

What Israel required was America to do the heavy lift of occupation and boots on the ground. But there has been major pushback from the MAGA base about that.

So: What's been accomplished? Iran understands that using a proxy army to harry Israel cannot be done with impunity. Will Israel be satisfied with the current state of affairs vis a vis their security? I have no idea how much more Trump is willing to do. Obviously Israel would want an Iraq-style regime change and occupation but I don't know whether that would bode well with the November elections and the Republican prospects. The MAGA base is very much America First with both Israel and Ukraine (for example).

I read Richard Marcinko's leadership book (Marcinko was the SEAL who founded DEVGRU, the SEAL's most elite unit, aka Team Six). From it, I concluded this: Applying Navy SEAL principles to lead people works best when the people are physically and mentally built like Navy SEALs. Most people are not, not even elite company CEO's and their staff.

It becomes a game of square peg / round hole.

Your digital assets are safe, as long as the data centers exist.

Banks won the right to tie their deposit accounts to their riskier trading activities in 1999 with Graham Leach Bliley "Financial Services Modernization Act" which finally fully repealed Glass Steagall, which had been getting chipped away over the years.

So, they received bailout protection - "privatize the profits, socialize the losses" to protect the consumer deposit arms of their organizations, when the rest of it blew up.

Transferring cryptocurrency and other asset risk to retirement accounts has the same effect - de facto bailout protection via the public treasury.

It's all about transferring risk. It converts crypto from "tractors in a online game" to "tractors in an online game with de facto bailout protection."
 

If it's a cost-plus contact, the incentive is to milk it for as long as possible, without delivering a product. It's the definition of a perverse incentive.

After my failed upgrade experience, I realized that if one puts Windows on "unsupported hardware", it would be trivially easy for Microsoft to disable it if they felt it furthered their interests. I realized I didn't want that Sword of Damocles so am looking at Linux.

One has to know what's going on - your disk is being encrypted by default, and all your data is getting sucked up to Microsoft's server by default. Your medical information, tax information, etc (i.e. PII and PHI).

Then, one has to research how to disable all that stuff.

Out of the box, by default, for the average user, all of the data theft and forced-encryption is going to happen because they don't know what's going on. The encryption is secret and the data theft is not advertised - just that one needs a Microsoft account to use this operating system.

I accidentally bricked a PC with Windows 10 Pro because I tried to upgrade it to Windows 11 using a popular third party tool to run on unsupported hardware, and selected the wrong parameters. I'm looking at the PC now and it's solid hardware and I don't want to send it to the landfill because Microsoft and its partners decided they wanted a windfall profit. I realized finally that if I knuckle under, I am voting for Microsoft and partners to only get stronger and bolder. So, I am finally going to put Linux on it.

Linux is a hobbyist's tool. You have to know how to secure it and how to update it yourself. I've used it extensively in the past at work. Microsoft Office and other programs I use don't run on it. It will have a learning curve. I can only do it because I am a professional programmer with significant Linux experience which is a small group. It will have zero impact on Microsoft, but for me, it is the Right Thing To Do because the only way to avoid this expensive and outrageous corporate control and privacy invasion, is to do it.

What about for people with no intention or ability to take delivery?

"a new breed of AI systems that understand the world, have persistent memory, can reason and plan, and are controllable and safe,"

Pick three.