nicpottier

I agree that HN is still pretty liberal but I also agree with GP that the swing to the right in the past few years has been very real.

Ooof. Maybe as a first pass it should check for people in the photo and just stop if there are any.

It doesn't just do rooms! Hilarious tweaking of a person too: https://imgur.com/a/1Quv4MI

This is wild, do you mind sharing what the process is in broad terms? I'm super impressed!

This is a neat take for sure, will have to play with it. Go is so great for APIs but really nobody has built anything that is anywhere near as good as say Django for general websites. I will admit I'm a bit skeptical that this approach doesn't follow say the Django approach of application/router/view a bit more closely as I think that's about as good a setup as I have seen but will reserve judgement until I play with it more. Either way kudos for doing this!

This is clever but isn't this dramatically easier than actually doing something with the input? You've gated things (and though I didn't crack it I'm really not convinced it is secure) but you also aren't DOING anything. That's a much easier problem. There are easier and more effective ways of gating LLMs based on a passcode.

Can you formulate a few for us and we can try them out?

I think your article is a worthy read and has some fun examples. I'm not so sure I agree with your conclusion though. The limitations you pose both in your article and here are real but already have answers in some cases. For one equipping GPT with upgrades for various tasks helps tremendously, IE, Wolfram Alpha for math, wikipedia to refresh its memory on facts etc.. I think you pointed out a really interesting edge case for the goat puzzle but also showed that with different words it CAN reason correctly but trusts it's memory more (and I correctly).

I think you are setting far too high a bar with "majority of information worker tasks". My wild ass guess is that with some thoughtful design and focus as it stands today GPT-4 could do 25% of tasks (or make workers 25% more efficient). The economic impacts of that are vast and I'm not particularly optimistic about who will benefit most from them. That's what the fuss is about IMO, not that we have some kind of AGI on our hands, which is far too high a bar.

Looks like this is 12b parameters. Will this fit on a 32gb M1?

And my reply would be that's because you likely don't know much about plumbing.

I'm not sure what we are debating here. On one hand you said your prediction was fantastical but then you are trying to defend it? Plumbing and bookshelves are really different things. One has to be built in place and is completely custom, the other is a consumer item. You can draw parallels if you want but it doesn't make it a strong argument.

> Doing anything beyond that is not only a bad solution, it is immoral to press your hand down on the weight of one candidate over another based on his or her identity categories. It is as wrong to prefer a white candidate as it is to prefer a black candidate.

This is a very convenient position to take for white men in power. Now we insist it must be a perfectly level playing field, never mind the centuries of racism.

My view is WE are society, change starts at home etc.. Again, my point here was that pointing out lack of diversity is absolutely appropriate. I disagree that we get to just say it's the pipeline's fault and not try to do better in our organizations.

The world of bits and bytes is a lot easier and faster to change than the world of hammers and nails.

Yes, right, so much free time to pursue your passions when you are unemployed.

Having some compassion for people losing their jobs isn't uncalled for. Especially when the same is going to start happening with programmers soon.

No. This isn't going to happen for any number of reasons. Your example doesn't even work as plumbing is still a manual process even in printed houses. (which are not going to be a thing either) GP is right that at least in the US the trades are a pretty secure job.

Yes, you don't want inclusion to be your only bar, but if you don't state it as a goal from the beginning then it won't even be considered. It is becoming a bigger and bigger red flag for an organization to not have some diversity of gender or race, pointing it out is not inappropriate.

I drove from Seattle to Denver with no planning (Tesla) and it was fine? It added about four hours to the trip as I recall but the car told me where to stop and charge etc..

EVs aren't the best road trip vehicle but that's ok, I don't road trip every day. They absolutely are the best 99% daily vehicle though if you have home charging.

Nothing is the best at everything, the EV compromise is pretty darn acceptable though. (that said, PHEVs with 80 miles of range would be even better)

Same, though I don't know why, GP matches my understanding as well.

Are these hot takes a thing on HN now? I hope not. Write a blog post articulating your argument and link that, but please don't turn HN into Reddit any more than it already has.

So they say, and I agree this seems likely. But until they release it that's all talk, and they have plenty of incentive to release it.

Until we see otherwise, don't we have to assume there's some secret sauce? Bard doesn't match GPT4 and it isn't for a lack of trying. (though perhaps that will change, so far that's the case)

Try making bad guesses and see it fall apart. (IE, guess something higher than your last guess when it says "too high") It is fully dependent on your part in the charade to keep the act up.

The hypothesis that I find most compelling and intuitive is that language is thought and vice versa. We made a thing really good at language and it turns out that's also pretty good at thought.

One possible conclusion might be that the only thing keeping GPT algos from going full AGI is a loop and small context windows.

Hah, hadn't thought of this but kind of love that take!

I thought your "guessing game" question was an interesting one so tried it on GPT-4. In my first attempt I played logically and it did fine and I finally guessed correct. On my second I made suboptimal guesses and it didn't stay consistent.

The thing to remember is that GPT has no state apart from the context, so it can't "remember" anything apart from what's in the text. That doesn't mean it shouldn't be able to stay consistent in a guessing game but it does mean it can't keep secrets.

Some of that can be solved with layers above GPT where say it it told it can save "state" that isn't passed on to the human but fed back in to generate the next response. But the size of that context is very limited. (a few thousand words)

There seem to be a fair number of experiments playing with giving GPT this kind of long term memory, having it establish goals then calling it over and over as it accomplishes subgoals to try to work around those limitations.

I'm guessing they mean more it's a limitation of where the cameras are placed rather than the processing power. HW4 slightly tweaks the camera positions and angles which I believe removes the current blindspots that make birds-eye impossible.

This is a great reminder, thank you.

As someone not skilled in this art, is there anything preventing us from opening that context window many orders of magnitude? What happens then? And what happens if it is then "thinking in text" faster than we can read them? (with an intent towards paper clips)

This is a genuine question, I'm not trolling.

I mean saying that using ChatGPT will make you one of the most elite coders in the world is a pretty high bar no?

I have been skeptical about all this, but decided to play with using ChatGPT for coding a task I had. It wasn't perfect, either in workflow or in final results but what was clear was that not adopting it will very quickly make you lag behind your peers.

I think the additive bonus is especially high for senior devs who are fast, in that those will be faster in grokking the generated code and better at guiding it to the right solution.

I guarantee you that in five years anybody who cares at all about their productivity will be using it constantly in their workflow, probably much sooner.

Are you a machine?