I don't think that the rich have such a plan. Rich just try to stay rich and do their moves accordingly.

In theory there is a solution. Governments need to buy or build companies that fill basic needs for the citizens. When you have that, you can use mandatory rotating employment in those companies to keep stuff running without paying wages and the companies themselves will provide basic needs for everyone for free. If automation replaces workers, it will have no effect to the country, except people get more free time. This is the only long term solution that I know of.

For short term solutions:
- Don't tax pay-checks
- Increase corporate taxes
- Increase progression in wealth taxes.
- Transfer all worker related requirements from company to the government to lower risk and cost of employing someone.
Basically try to make human labor as cheap as possible. This will not work in long term, but it might help short term.

"Why Democracy Is Mathematically Impossible" explained by Veritasium: https://www.youtube.com/watch?...

It is funny to you only because you are not looking the big picture.

"As Amazon has grown, the number of independent businesses has fallen. Between 2007 and 2017, the number of small retailers fell by 65,000.[8] About 40 percent of the nation’s small apparel, toy, and sporting goods makers disappeared, along with about one-third of small book publishers."
https://ilsr.org/articles/fact...

And that is just job loss inside the USA. Amazon also sells outside of the USA so similar job loss happens globally also. So Amazon is currently eating away their competitors and employee count is growing because their market share is growing faster than automation can replace workers. Once they can no longer grow, automation will just decrease the employee amount.

Inventing new jobs is extremely easy. Just give a spoon everyone and ask them to dig a big hole on the ground. When it is done, ask them to fill it and repeat. Real problem is finding someone who will pay for work.

So you should not think what new jobs we will have, you should think about what people want to buy and who is going to pay for it and how much. Humans have some basic needs, like food, but that is already heavily automated. We won't be needing much more food so there won't be new jobs in the food sector, unless they eat old jobs away at the same time. Most likely the amount of work in the food sector will just decrease in the future as automation increases even more.

I think similar thing will happen with every need humans have. In the past 15 years, humans have lost steadily about 0.4% of population worth of work per year to automation. That is not because there is no work to be done, there is just no money to pay for it, because human labor is so expensive.

Personally I would rather take AI designed drug that is not tested in ANY way in any animal or human. Than take a drug that has gone through all required animal and human tests. But that is only because I know how they test them with the AI.

But luckily for you and perhaps for me also. They will have to run all the same animal and human tests for the new drugs, before they are taken into use. So these drugs will be at least as safe as any traditional drugs would be. But because AI can actually check the binding against ANY known human protein, they should be much safer than what you can cover in mandatory tests on humans.

I remember covid. But this method of making drugs is completely different compared to that.

Toxicity screening part became easy because AlphaFold2 created 3D model of all human proteins, so you just need to run the binding test against all of those to see if you have unwanted side effects or not. If the drug doesn't bind, it shouldn't in theory have any side effect.

> But "the bigger point is this: little by little, Apple has been laying the groundwork for its generative AI efforts with some pretty interesting and novel ideas."

"The adjective "novel" means new and original, not like anything seen or known before."

Interesting idea perhaps, not new nor original:
From 2023 "CodeFusion: A Pre-trained Diffusion Model for Code Generation"
https://huggingface.co/papers/...

Food causes 40% of deceases. There are some rare genes that can protect you from cancer and other deceases. If you want to live long your options are:
1) Get your genes checked, if you are lucky and have those genes, you have nothing to worry about. Or perhaps, if you had the genes and would eat healthy you would live even longer and more healthy.
2) Eat healthy.

Regarding the breathing, oxygen is toxic for all living things and it literally kills us if we get too much of it.
https://gladstone.org/news/res...

And even if have normal amount of oxygen in the air, consuming oxygen causes free radicals which damages cells and causes eventually death. Generally speaking, if you can slow down your breathing, you should live longer and stay younger, assuming you can still keep your cells operational. There are even studies that show that a lot of physical activity can lead into reduction of lived years.

No. Norway lets companies sell oil, but they tax those companies about 50% of it. But government doesn't spent that money instantly, instead they put that money into a fund.

This makes a big difference, because there are several countries in the history that got rich with the oil, but they either let the rich or the companies to have it all, or they spent it all instead of securing their future.

> And how many of those applications allow for a human being to be removed from the workforce?

Oh, I did not try to argue with you. I just assumed, based on your first line that you were not aware of all the things that AI is being used in the corporate world. But bottom two links are cases where workers are being removed. And it is not like this was a full list of what AI has done, it was just a few examples. If you were not familiar with any of these, I can guarantee that you are not aware of hundreds of other examples there are.

> it looks like it will take highly technical staff just to have the capability to review the AI returns to confirm if they're even workable or not

Yes. If you (or the AI you are using) find a new mathematical solution, you will need a mathematician to verify the result. Then you need some kind of circuit engineer to implement that algorithm into a chip. After that consumer can just buy those chips and take advantage of faster computation power, or in this case, the company can get the benefit of selling and using better TPUs. But while you need a human, you only need it once. Once the new algorithm has been verified and taken into use, that human work is no longer needed.

> so far it doesn't look like the people who fall into the vast chasm of nontechnical "office workers" are under immediate threat from AI.

What does a nontechnical office worker do? Journalists? Some of those tasks have already been to AI or even simpler algorithm. Duolingo at least was going to replace some tasks with AI. But generally problem with these non-technical office jobs is that they are often easily replaced with a simple script. I personally once created a script in week, which replaced a full time worker, who was later fired as there was no other work left for him to do. And I know people who use an external calculator to count the sum of values in their Excel worksheet, every day.

You could probably count the sum in Excel already decades ago. Yet some people still do "manual" calculations, simply because they don't know how to use their tools. So it is not like we need a new technology to replace people.

So I predict that this is not how AI will replace half of the population. Instead, AI will make scientific discoveries, like invent cure for everything and this in turn, will cause massive job loss for doctors and nurses.

> To date the only AI that I've seen deliver any sort of semi-useful work

I hope that you mean that it has been the only example you have personally seen, because there are dozens of examples.

"AlphaEvolve proposed a Verilog rewrite that removed unnecessary bits in a key, highly optimized arithmetic circuit for matrix multiplication. Crucially, the proposal must pass robust verification methods to confirm that the modified circuit maintains functional correctness. This proposal was integrated into an upcoming Tensor Processing Unit (TPU)"
https://deepmind.google/discov...

"AlphaEvolve is accelerating AI performance and research velocity. By finding smarter ways to divide a large matrix multiplication operation into more manageable subproblems, it sped up this vital kernel in Gemini’s architecture by 23%, leading to a 1% reduction in Gemini's training time. "
https://deepmind.google/discov...

"AlphaEvolve discovered a simple yet remarkably effective heuristic to help Borg orchestrate Google's vast data centers more efficiently. This solution, now in production for over a year, continuously recovers, on average, 0.7% of Google’s worldwide compute resources. "
https://deepmind.google/discov...

"On Wednesday, Google said it had proved it could cut total energy use at its data centres by 15% by deploying machine learning from DeepMind"
https://www.theguardian.com/en...

"How a Japanese cucumber farmer is using deep learning"
https://cloud.google.com/blog/...

"how law firms are using AI to remain competitive"
https://www.thelawyerportal.co...

It is not a paradox. Productivity has increased at least since 1945 and it even accelerated around 1995. But benefits are not distributed evenly.
https://i.sstatic.net/iCTuo.jp...

This productivity increase, or automation, has also caused a steady job loss (when measured as total hours or work), at least for the past 15 years. I am pretty sure that this means that even if new jobs are created, it won't be enough to cover lost jobs, unless we switch back to "man works, woman stays home" society.

You are also mixing up AI and LLM. LLM is just one kind of AI. For example Alphafold2 is not an LLM and it has already done a job that was worth about 100 000 000 years of work. And it was not just any kind of work, it was extremely important work which humans were ready to pay on average about 100 000 dollars per year. Even simply AI that just sorts images into two boxes can automate several different human tasks from farmer to doctor. But most of this is still unimplemented.

Regarding your B. When you make decisions in a meeting, it can help years later if you can check why the decisions were made, when there is a need to change those decisions and you don't remember why it was made. It can also save your skin if there is something written down that says that customer made a wrong decisions years ago, so you won't get blamed for it. It can also help when people in the project change.

There are 2 kinds of meetings:
1. Workshops, where you try for example to find a solution to a problem and the solution requires multiple skills that are not found from one person. I often have these with my colleague when I know one part of the code and he knows the second part and we try to decide which technical solution we will use for our problem. I also use these with the customer when I know how computers work and they know the business logic.
2. Information sharing. These meetings could almost always be just an email. There is usually some time in the end where you can ask questions, but most of the questions could be asked via email also. These meetings have often hundreds of people which makes it extremely inefficient. Big bosses like these kinds of meetings.
3. Status meetings. These are meetings where some kind of manager calls the meetings and then asks people one by one to give their status report. These meetings however are not about the status of the project, they are about the status of the manager. Employees themselves don't usually find any value for these meetings as they often take an hour or two of their time per meeting. You can replace these with scrum-dailys that take just few minutes and gives all the info you need to keep up with the status and spot problems early. Status meeting is not a real meeting type. It belongs to information sharing category, but I listed it as a separate type, because this is so commonly used.

Most common mistake is to mix information sharing and workshop into the same meeting and invite people who don't participate into the workshop, but they listen anyway "to get the information". Correct way is to have a separate workshop, write the decisions down and then send it as an email and if needed, have another meeting after that for any open issues.

You can get AI generated face images easily from the net, so what is the problem?

Generally speaking, AI is bad for marketing, because humans favor human made products even when they would rate the AI made products better than human made[1]. There are some exceptions to this, but those are quite minor. There are some reasons to this, but I think the most common reason is that humans don't understand how much better the AI is compared to humans in some tasks. I have talked to several people and when ever I bring up the fact that AI performs better than humans, they are shocked and demand evidence. Other common reason people say is that they want to favor human work (but it is hard to prove is this their real motive).

Obviously if AI is worse, you get rejected for a good reason. My point is that even if your AI is better than humans, you will still get rejected, because of the bias humans have.

[1]
66.7% preferred AI made music, when they didn't know it was made by AI and 11.1% preferred human made (which shows that AI made music was better).
80% preferred human made music, when they were told it was made by human, no-one liked the AI-made music when they knew it was AI-made.
https://essay.utwente.nl/96214...