Why is it a web client? I’d much rather have a local client.

I’m 90% certain that none of the resources being “bought” by AI actually have been sold, or created yet.

Indeed. The prices skyrocketed because vendors realised they couldn’t get replacement supply in the future. What existed today was all they were going to get.

I’m expecting a glut of supply once those contracts fall through.

Speaking of the auto industry: Is AI a mirage, like the dreams of a working rotary engine?

It is, but I think it’s a different type of mirage. The rotary engine does work, but it brings with it significant downsides. Getting the positives without the negatives is the mirage being chased.

AI appears to do one thing, but actually does another. People see it “creating” new things, but it’s more like it shreds work up and then glues the pieces together making sure it looks consistent. Train it on one work and it can reproduce that work. Train it on two and it will mash the two. Train it on a billion and it will mash the billion. Nothing creative,. No extrapolation. Just interpolation.

People want the AI promise regardless of the downsides. It just doesn’t exist.

I work in the semiconductor industry, but have little to do with physical aspects.

Ok. I’m not familiar with them. I was just relaying my impression from the start of the article.

Article is about research by IMEC (a chip packaging company) into dealing with the heat problems encountered with stacked dies such as HBM memories with a GPU.

You have to get a third of the way in before they really get to the topic.

Well I don’t think they’re growing. Depending on the sector all the AI data scanning is a concern for a number of companies. I’ve seen a lot of European discussions about getting away from US cloud services too.

If they lose all the money they have pumped into AI, then they will be relying on Windows and Office.

Good for them that both of those are currently doing fine.

It’s a big matrix multiplier that is tailored for machine learning model evaluation (not training). Often they are low precision as that’s all you need for model evaluation (or “inference”).

Think of it as a much less useful GPU because it won’t do graphics.

AI code sucks. An AI code review tool told us.

There’s a lot of crossover here. It encompasses the “you rent everything and own nothing” aspect of today’s society. It’s multi-billion corporations against consumers.

A win here will be something that will expose chinks in the armour for a lot of other issues. I know it might seem unimportant because it’s games, but that’s its beauty. It’s a backdoor into so many other fights.

I’m not the only one seeing the 20 or so deals like this in the last few weeks. Right?

Is this because the SEC is shutdown?

This feels we’re going to get a movie like The Big Short in a few years time. “The Big Hallucination” or something.

While the drones are armed, they use non-lethal or less-lethal weaponry,

Less lethal? So still lethal.

Why is everyone so keen on making Terminator real?

A better way to say that (more accurate) is that abusers often learnt “how” through personal experience.

The key point is not to look at someone who has been abused and think that they are likely going to become an abuser.

They maintained it, but I think it was energy negative.

That’s obviously a cello.

“Recieving stolen goods” is prosecutable.

It’s a lesser crime than the original theft though.

ID seems to be quite ingrained into Spanish life already. It really surprised me when I needed an ID number to buy a metro ticket.

They are under development, and there is a small market for development machines. It also allows the manufacturer to understand the issues they’ll get once the high performance processors are here.

Nobody has really made performance implementations yet. They’re all IoT level or low end mobile device.

Tenstorrent are probably the closest to having something serious.

If this is is the guy that recently had Bernie on, then I think he’s mainly poorly informed and unable to realise when a politician is spinning a line for ulterior motives.

Ah, so you’re a waffle guy!

https://m.youtube.com/watch?v=LRq_SAuQDec

The AI bubble is currently grinding my gears on this. “XXX is an open source model”. No, it’s not. Do I have access to all of the information necessary to recreate it? No, I don’t as nobody releases training data.

Training data is the source of these models. Without it, they are just free use.

The trouble is that “core” is just that. The heart of the processor. There’s a lot of shared state in the caches and the TLBs which is all common to multiple cores.

At this point I think speculation attacks are almost being accepted as the price of having high performance processors. It’s almost impossible to rewind all non-architectural state when you hit a mis-speculated branch.

Probably because although there are fabs going up around the world (USA and Europe) TSMC Taiwan seem to hold the latest technology nodes, and aren’t they interested in growing capacity. They seem to like having the high end expensive limited process. All the other fabs are coming up with processes 2 or 3 generations back. (5 or 7, not 2 or 3).

All means that although there’s a market for the optics, it’s not the bleeding edge stuff.

“Dozens”. Really?!?!

With batteries that would have a multi-day cycle like these ones, you’re going to be trying to flatten out the demand curve (and supply, but the two are related).

The US generates 4.2 PWh a year, and so averages a consumption rate of about 480GW. So, in an ideal system we’d only need this level of generation capacity and if it was higher sometimes and lower others the batteries would smooth it all out.

I’m going to take your 560GW figure as representative of normal demand above the 480GW average. I’ll say half of every day is 80GW above average (when we’d be draining batteries) and half is 80GW below (when we’d be charging). The real curves are much more nuanced, but we’re establishing context. 80GW for 12 hours is 960GWh, so let’s call it 1TWh of battery capacity needed for the whole USA to smooth out a day.

That’s 117 of these installation, which frankly I find amazing that it’s so low.

I don’t seen how else you do it.

“Removing the stigma” is desensitizing by definition. So you want to desensitize through… what? Education?

Yeah I mean it’s just a more easy to use Photoshop basically.

Photoshop has the same technology baked into it now. Sure, it has “safeguards” so it may not generate nudes, but it would have no trouble depicting someone “having dinner with Bill Cosby” or whatever you feel is reputation destroying.

Technically and legally the photos would be considered child porn

I don’t think that has been tested in court. It would be a reasonable legal argument to say that the image isn’t a photo of anyone. It doesn’t depict reality, so it can’t depict anyone.

I think at best you can argue it’s a form of photo manipulation, and the intent is to create a false impression about someone. A form of image based libel, but I don’t think that’s currently a legal concept. It’s also a concept where you would have to protect works of fiction otherwise you’ve just made the visual effects industry illegal if you’re not careful.

In fact, that raises an interesting simily. We do not allow animals to be abused, but we allow images of animal abuse in films as long as they are faked. We allow images of human physical abuse as long as they are faked. Children are often in horror films, and creating the images we see is very strictly managed so that the child actor is not exposed to anything that could distress them. The resulting “works of art” are not under such limitations as far as I’m aware.

What’s the line here? Parental consent? I think that could lead to some very concerning outcomes. We all know abusive parents exist.

I say all of this, not because I want to defend anyone, but because I think we’re about to set some really bad legal precidents if we’re not careful. Ones that will potentially do a lot of harm. Personally, I don’t think the concept of any image, or any other piece of data, being illegal holds water. Police people’s actions, not data.

Why is the cheque redacted?

Who and what is being protected?

Maybe not peaked in terms of performance, but in terms of rate of development … Absolutely.

Slashdot.org and digg.com

I’m old school.

Our new moral guides.

Techno-priests

We have a fairly big step up in pay from junior to senior. I can take on 2 or 3 juniors to a high senior or especially principle engineer.

We’re often also taking juniors on that we already have a relationship with through placements during their university course. That minimises the risk.

juniors are a way bigger risk than seniors and usually leave a company right around the time that they’re getting good.

Personally, as a manager, I find the opposite.

It’s always the juniors that exceed expectations. You never hire somebody senior and find they can do twice as much as you thought. Juniors are often eager to learn if you are willing to teach them. They want to be good at their job, because they know they are laying the foundation of their career. Seniors often have all the bad habits baked in.

Then, if you get a good reputation for developing people (because they leave your team and impress their next set of colleagues) it becomes easier and easier to hire.

The problem is the first part of your question.

You only need Bayesian filters for that IMHO.

Oh yes, mgh not mh. You’re right