jacquesm

That video is something else, wow.

Another thread like that:

https://news.ycombinator.com/item?id=1319995

Gah this can be so annoying. I've got one going for 56 years. A cassette tape that I had as a kid, all I remember is that it was orange and that it had the most amazing jazzy tracks on it (mostly instrumental). I've listened to every orange labelled tape on various tape archives and still haven't found it, the memory has faded far enough that I only occasionally remember snatches of it. But I'll keep looking.

Other ones: Tony Banks, Portrait of Jerry Jackson

https://www.youtube.com/watch?v=n_U0jhtzmzI

Cori Josias, Takin' it straight

https://www.youtube.com/watch?v=DPuCra-t_Bw

It slows down to the point I find it unusable so I ask it to summarize as one cut-and-pastable block, copy the block, open a new session and kill the old one. This usually happens after a few hours, probably because 'thinking tokens' even if they are not displayed crowd the context window.

At that time running that many concurrent tasks on such anemic hardware was a small miracle. This wasn't your average linux kernel.

It's a bit like having a metal layer of a chip and nothing else.

A connectome describes the connections between neurons. The weights are the degree to which the neurons are biased to fire or not and what you did with your training was to create those biases based on the training data. The actual weights are not in the connectome it is - as the name suggests - just a map. I wonder what the control experiment could look like, possibly a randomized connectome. And maybe it would work just as well as the fly brain. You can do the most horrible butchering of models and then train them back to functionality, this isn't all that different.

For me this works: "One thing at the time. This is a conversation, not a lecture, try to balance the amount of words you write versus the amount of words that I write. Don't overwhelm me with 10 pages of prose where a single sentence would suffice, brevity is a virtue, not a defect." I've had to tweak it a couple of times, probably because of different versions of ChatGPT but it is usually a variation on this that will do the trick. Besides being far more interactive it takes the frustration down quite a bit.

That's pretty much the Gemini experience right now. It is unusable. The funniest bit is that the LLM is not in on the joke so it will act like it never happened... Google once again wrecking their own products.

This is not that.

Stalker. It was unexpected but not for the stalker. What a creep.

It's the opposite.

Get him to remember to take his keys.

Like they do with the homeless today you mean?

That you need it to survive.

Of course. But that's not a license to ignore reality.

Yes, this is the perfect time to do so, after all, what could possibly go wrong?

Really, there is a contingent on HN that seems utterly divorced from reality. The trucking industry as a whole profits ~30B annually in the United States. That is an absolutely insane number of miles hauled. Many of these operations are barely hanging on and whether or not 16 trucking companies going bankrupt in a month due to fuel prices is normal or not I couldn't tell you but on the off chance that it is this might not be the time to experiment with what other things you could do to that industry.

For contrast: Amazon profits number in the hundreds of billions (300B+ iirc) and that is just one company vs a whole industry.

Effectively that means that they are making pennies of profit per mile driven and that any kind of destabilization of that industry will instantly translate into either shortages or further price increases.

It's a war. What you call it is immaterial, just like it doesn't matter that Putin refuses to call the invasion of Ukraine a war. The identity of a thing is not dependent on the label, it is inherent in the thing itself.

Yes, let's cause the last lifeline that keeps society afloat to fail.

This is all very nice in theory but you do realize that if US nationwide trucking is no longer able to operate profitably (and they are barely profitable as it is, see TFA for a nice illustration) that you are looking at absolute chaos in short order.

You'll have some nice and undamaged roads to look at though, but I'm not sure if that will make up for the downside.

Pretty much. And, interestingly, even though it is a little bit slower than most other models I can run here it is quite a bit more efficient with the number of tokens used so on longer runs it often comes out ahead in terms of wall clock time. YMMV of course this is for pretty specific workloads and there is a good chance that on other workloads it would perform worse. But so far it has been a pretty good model for us.

Because you need to do both. Recording creates material that can be damaging and distribution realizes that potential. Banning the first is more effective, banning the second should come with additional penalties because it proves you did the first and then went on to the second presumably willfully ignoring two different legal aspects, creation and distribution.

Great work, keep at it!

Strawman detected.

Sure. But the ability to do this kind of rugpull is new and so you will get responses like these and that's perfectly logical regardless of the legalities. I'm not saying you can't. But I am saying that if you ignore your users - even if what you ship is free and open - you are still going to be subject to backlash if you change course very rapidly. You could also simply create a new project and support the old one during some transition period. Principle of least surprise and all that.

Another thing is that the Black Plague symptoms are pretty horrible, enough to cause anxiety even if you know there is a cure.

Ah, but then whose time would that be? How do you keep track of what solar time was parallel to 'your' time?

Time is such a fickle thing. Every atom in the universe knows what time it is, but between the effects of gravity, the speed of light and some quantum stuff it is amazing that we can agree on this at any level at all. I spent a lot of time earlier this year getting some ultra stable clocks to run without shelling out 5 digits and it was a very interesting little detour. Eventually I settled on an old atomic clock module from a decommissioned cell tower, not quite 'miniature' but small enough (I've seen some of the newer ones and they are incredible). The thing is precise enough to show the day/night deviation in phase from DCF77 to my home a few hundred km away.

What makes you so confident? Do you have any proof of this? If so I would love to see it because it would explain some stuff I've come across, but could not place.

I can't really agree with that. If you ship your linker with a disclaimer saying 'don't use this for early stage stuff, one day we might decide to rewrite the whole codebase in a language that won't be available early enough' then you'd have a point. But if you didn't that is precisely the kind of use case where mold shines, and so inevitably it will be adopted. Your downstream should be precious. At a minimum you should communicate such intentions, do so timely and hear what others have to say, even if afterwards you decide to push on.

They were solved by mold. And then mold decided to un-solve them.

The weird thing is that a lot of people that are ultimately successful share a lot of these traits. But I think the big differentiator is in how you deal with any of them, you can have the trait of not immediately rolling over when an expert says something can't be done. That's fine. But if ultimately you can't do it either you simply admit the experts were right after all and you move on. The further down the list you go the bigger the success if you eventually do manage to deliver, but the bigger the loss if you don't.

Where Cringely went from interesting to a risk to himself and others is where he refused to stay connected to reality.

I can relate: I start a lot of projects where I'm usually in over my head. But I always learn, sometimes learn a lot and some of those projects eventually do work out. But I don't have my ego invested in any project and I think that is one of the bigger differences, some stuff works and is obvious, some stuff could work and is not obvious, some stuff can't work but it is not obvious (to me at least) and some stuff simply can't work and it is obvious. The third category is where you lose the money, the second is where you make it, the dividing line between the two will give you Theranos, UBeam and Moller on one side of the line and CISCO on the other.

Yes, really. But he had a lot of people that took it as gospel and he did motivate a lot of people to go into tech. Mixed bag.