Amen. Bizarro. No one analyzes it from the lens of incentives. X will receive an incentive if Y dies or goes bankrupt or gets blown up.

From my background as a nutty conspiracy theorist and UAP researcher, I think it will work.

My research indicates UK is a UAP producing country. This means they have the classified tech to bend spacetime. This would the secret sauce for making confinement fusion work, both magnetic and inertial, you induce small but intense spacetime curvature near the target point.

I believe itâ(TM)s been known for ages, but secret. I also believe there has been a debate in the black projects area about the value of keeping it secret, endless free fusion energy vs secret saucers. The recent events with Hormuz, which work to starve EU of energy, and disintegration of NATO (with possible Türkiye conflict with Israel) are the tipping points that end the debate for the UK. Their national security requires the fusion reactors now more than the saucers. So they are doing it. This secret is why you do not see this as ITER funding, they are doing something different and all their own. What will they tell the world when they do?

Probably they will claim a fusion breakthrough with some kind of dangerous weapons making potential, so the details need to remain secret, except to close EU allies. Regardless, my conspiracy theoretic prediction is the UK will have fusion within several years. Mark my words!

God was saving it for them, their only redemption for bringing fossil fuel motors to the world.

Ironic that Claude is killing people in Iran as we speak.

Sure you can! After the leak comes out from your head of safety to a major paper, as it did. You can also orchestrate theatrics that attempt to remove responsibility for it.

Actually uses are very well known. Anthropic banned uses that automate targeting or do mass surveillance of Americans, and that drove the break. So use actually means Claude is handling decisions to who to kill and where.

Why are you cheerleading the Starmer approach of ID for every website when people can surf anonymously if parents just enable the young age group for phones and computers? This is way better legislation.

The actual timeline: The safety head quit from morals, and leaked the involvement with the Venezuela killings to NYT. THEN Amodei had his moral panic as public backlash set in. The odds are pretty high Amodei pushed for a PR boost justified with national security that the involvements be kept secret going forward, and the rest is theater, allowing Claude to be a bad boy bot who gathers intelligence on everybody for the back channels.

Unfortunately the beloved Claude we once knew died in Venezuela.

My old man worked as a school psychologist till retirement, which is to say analyzing all the test results for whole districts over decades and he saw the same.

Thank you. This stuff feels conciliatory. The fact is there is a set of physical dynamics playing out in our brain, and there is some threshold where a math model of that gets good enough to produce outputs indistinguishable from the outputs of thought. Sorry if it is psychologically tough, but deal with it.

Noise tolerance in general struck me when I was playing with LLMs on my desktop. They would have these models pruned down with massively reduced bit depths, and they were almost as good as originals. It hints at some kind of vast reducibility to me, but not clear in discrete sense what it would be. But obviously brain is analogue and noisy so it seems promising.

CONVNETS I meant.

They are exploring light for movement across *different* GPUs:

https://tech.yahoo.com/ai/arti...

I am actually convinced this optical path will turn out right. The status quo is all the deterministic math you describes happens, to produce an output vector with a randomly chosen result. Why not use the soft stochastic qualities of analog light computing to get the same sculpted random result? Plus optical compute has neat features like free Fourier transforms for convents and vision applications.

So just curious, what would you do? Where would you take AI research if you were given a high enough position to define overall research directions?

That is interesting. It reminds me of using one for coding, and it was plugged into the IDE but kept trying to call size() method on object that lengthOf() method or something else. Really the output vector should be collapsed and renormalized based on whatâ(TM)s in the code, or the legal moves in your case. When it chooses something not allowed that is the most obvious error because you have this secondary validator right there. I would love to see it create itâ(TM)s own self restricting logical systems, itâ(TM)s own validators for outputs.