RugnirViking

i'm sure jensen huang will be fine. He has one or two other accomplishments to his name

remember to test against benchmarks. I would love to hear about your progress.

Reckon you could give us some idea of the other side? What are the alternative explanations proposed? The passage here (albeit translated) seems pretty unequivocal.

It also doesn't seem that unlikely. I know a fair few adults who mutter to themselves under their breath while reading, never having managed to kick the habit of reading aloud.

about 3 years ago I randomly stayed in an airbnb in an apartment block in lisbon that had one ! Theyre still around in old buildings, it felt very sketchy but I got used to it after the first couple times

there are higher levels of writing than "grammatically correct"...

we don't dismiss claims of poorly written code with "but it compiles!"

same. we played 5-6 times I think in uni, but the peak for sure was a marathon two day session with six players. We had cocktail parties with a side table with meals + drinks for the diplomacy phases. Legendary memory, though I also remember every time thinking "that was incredible but I am NEVER playing this again" each time afterwards before going back for more a few months later.

I've long dreamed of a reprise if I ever get the space and friends again, perhaps with dressing up fancy and/or a really nice champagne for that real diplomacy feel

absolutely not ! it's not an "ohhh ill get 100x more coding done" it's definitely a preference/work style thing. I find it's hard to get in "the flow" when managing multiple agents, and if I just do one at a time with today's frontier models, I find myself waiting 10 minutes twiddling my thumbs while they do something all the time. Then needing to catch myself and turn back over to it when its done, all these micro switches between tasks is really difficult for me. If I had an instant agent I probably would be somewhat faster, but not zomg1000xunicornrockstar nonsense. I would be way happier though

> Nobody needs to code at 600 words per second.

I do. I used to use haiku for the speed. Now its just as slow as the rest. Speed is my #1 ranking of how good a model is

is this a comparatively huge or a comparatively tiny amount, when taken with the money made from these activities? does it escalate if no improvements are made?

I ask because I dont know. The devil is in the details with this stuff

this. My not-at-all technical partner often comments to me that whatever youtuber is reading from an ai script and just turns off whatever we're watching. I assume similar happens in other areas.

For her its rarely the specific tells ("not-x-but-y" etc) but rather its bad metaphors and analogies that get her.

for what its worth, I was amazed when I moved from the UK to denmark that young people smoked in denmark. For me it was a thing for old people, mostly due to the price I suspect. It just wasn't cool - although vaping has changed both of these since I left, by my understanding

according to their annoucement for sol-6 and luna-6, they have also tried to improve responses, and the release post has examples of old and new language on the same task to illustrate this

surprising it took until xhigh before it did the legs properly - all the ones before that have both legs on the near side of the bike.

Absolutely. Its also an arms race - if one company advertised like this, maybe they'd get an advantage. But given that everyone does it, they're back to where they started, relative to each other (and also relative to the wider economy!)

Interesting. Wonder why its also shifted to opt/alt + k instead of j

absolubtely. Any random junior consultant can tell you what the book tells you you should do. If you want to actually do anything worth doing, you need to step beyond that in a few, limited areas, and follow convention everywhere else. Which areas? pay a senior engineer and they'll find them.

are the things that you linked to adverts? as in, someone paid facebook to promote them in the ad bit? or are they just regular content.

the law came into effect due to widespread advertising campaigns on facebook by someone (often thought to be russian government based) for both sides of any culture war topics, presumably to drive division. Things like simultaneously advertising BLM/ACAB + thin blue line.

ive been taking a look there. I think for some of the image to image ones you really ought to use real images as the base prompt, there are a lot of weird things happening where the base image has issues and then its hard to say if the model should be correcting the flaws or not. Things like "childrens drawing app where each crayon is clearly sized to be tapped on" but the base image has the rainbow cut off partway across with only half the red and black crayons visible.

Also "vintage" photography being corrected, where the original is clearly ai generated with unrealistic sharp focus everywhere

no it wouldn't, such things are straight up illegal in most places? You have to clearly disclose ads + have clear distinction between ad content and rest of content. A bunch of youtubers have gotten in trouble for this?

> Most bugs won’t be “coding” bugs. They’ll be “you asked for the wrong thing” bugs.

that was pretty much what coding bugs were already. Type systems and compilers sorted the other kind already. Unit tests and code review the rest

> Tokens are the new computing paradigm. Everything will be re-made on top of it.

what?

> In 99% of software it does not matter that you could’ve written a faster algorithm

a 0.5 second delay makes a huge difference. So does 0.1. Try setting animation speed to slow on your phone... You'll be tearing your hair out immediately.

you speak as though performance can be sorted out after you have users. Performance is extremely often a huge barrier to getting any in the first place. Jim from hyderabad doesnt care that it works on your iphone 18, hes using your software over a 3g internet connection.

Sure if you specifically prompt agents about such concerns, and specifically tell them that they need to make before and after timing tests, they will do things. Maybe even passable things. But my PM isn't going to do that.

This goes for a bunch of what coding is, professionally. My PM isn't going to ask if the fact the robot chose to use http in their prototype is a problem. They'll just deploy it right to prod.

My PM isn't thinking about if our interface is compliant to all the standards required. They also aren't thinking about scalability.

They aren't thinking about all the platforms we deploy to, and the quirks. They aren't thinking about the cloud costs, or any other costs we might employ. They aren't thinking about if the amount of network traffic their prototype is generating could be a problem for the services it talks to.

and to be fair to my pm, if I asked the LLM for what I wanted to use, it probably would be a niche product, and certainly not incorporate what the sales team needs to sell the latvian government on the next big deal.

You need specialists. "Taste" exists in the domain of technical decisions as well.

Some people think we will escape this with a massive long prompt telling an agent to call a bunch of subagents to think about all of this, and it will be instant and remember it all perfectly. That you can have a "dark factory" software company run entirely by agents (or more specifically, a company run entirely by agents apart from whatever their own job role is because of course they will be necessary unlike all the other people doing presumably very easy jobs I wont check)

Maybe it will, maybe it won't. Such things certainly don't work right now. They just can't contain enough context at once to do this (and much of the context is not digitalised in a high enough quality way i.e. without contradictions and errors). If we have a revolution where we get 100x context window or something we can come back to this conversation

looking at the mac keyboard im typing on, there is no key called "alt". There is one called option, and pressing opt+k gives me ∆.

Wonder why that changed? (or if it did?)

> let's get a nice prose out of it.

what do you think this process is? why don't we just write lists of ideas to each other, in the raw format you presumably started with? is there any value in it?

my opinion is that the process is literally nothing more than thinking deeply about the ideas and trying to phrase them in the clearest possible way such that you lead someone else into understanding them.

LLMs have awful, terrible judgement of what other people know, or are likely to infer from a piece of text.

> Nobody can afford not to use it

The games market has a massive oversupply. It's not enough to just "make a game", you have to make a good one. That is to say, better than all the other things your audience could be playing. Cheaping out is literally not an option. You can cheap out in one area to spend more effort in another area, but you can't just say "this is cheaper therefore it will be the default" because the default isn't good enough. Most games fail

> not really sure what it shows

it shows that you don't get anything for free. If it was without any cost in other areas to make a human massively more intelligent, it would presumably happen over time via the same processes that led to us.

As it is, the higher (and lower!) ends of the distribution tend to be highly correlated with other issues (mental+physical), and the further you go, the more unfortunate things which make it harder to do stuff in general start to crop up like psychosis, autism, ocd, anxiety, addiction etc

it would appear nobody here knows. We probably ought to shut up until a seismologist decides to chime in

Probably over twenty today in my work day alone. Its a normal part of working with agents. They do not always have the context they need, but fail to be aware of this.

Misinterpreting the results of performance testing it had just run to argue for the exact opposite of what the data it generated showed.

Stating something was current guidance (a quick read shows the document it found was from 2016, it was served the last edit date in it's API call, along with newer documents that contradict it).

Assuming what a Jira ticket said without ever reading it with it's tools and then doubling down on the contents it has assumed.

On learning specifically, it constantly gets grammar in foreign languages wrong it can write fluently when prompted correctly but when asked the sort of wrongheaded questions from flawed premises learners commonly make it is prone to making stuff up. I experienced this yesterday when asking about how danish comparisons work.

Heck i mean see https://alignment.openai.com/misalignment-reports/self-gener...

They are prone to make up restrictions for themselves you never asked for. Ive seen this behavior on occasion also.

Dead link?

> my mech eng cofounder has this "respect of the craft". From from my perspective this has only caused delays in generating revenue.

You need both. You really need both. A salesman to keep the craftsman from getting too bogged down in details to ever finish, and the craftsman to keep the salesman from ever lowering quality too much that customers start seeing the company as money grubbing misers trying to make a quick buck.

It only works with both. Too much of one and you sink.

This can be seen time and again in successful endeavours. (Apple, Microsoft, Google, star wars, honestly so many movies, moon landings, the list could go on)

> It distills the intuition of a whole world and is able to transmit it on demand, like in your examples. It's an interesting dichotomy.

The only problem I have at the moment is it's very prone to hallucinations, even now, yes frontier models astra fable with all the bells and whistles. This makes it hard to confidently use while learning because I have to be on the lookout for lies while I'm learning, which is precisely the moment I am least able to distinguish them. So instead theres just a constant low lying dread.

Nonetheless, I am able to get some value out of them. Just not all, everything requires manual effort to duplicate and check which you should probably be doing anyway as part of learning