digilypse

The point I got from the article was that AI is a cause for skill atrophy for any level of seniority, and tends to worsen over time. I can see this happening to myself in some areas, while in others I’m able to learn new skills I’m interested in.

Building a procedural open world at the scale of a flight sim in UE5, in a month wrote 5k prompts, spent 40B tokens and will likely take at least a trillion more over the next year of full time dev work.

Big solo projects exist, they're just not coming out as frequently because of, uh, how time works

they’ve been leaning into that since 2020, are the hosting platform for truth social etc

That makes much more sense! I was perhaps being uncharitable in my interpretation.

Does art lack meaning and purpose until it’s sold?

Your interpretation of retirement is probably more strict than most people’s, but even if it wasn’t, how can it be hard to imagine connecting with a community outside your coworkers and customers? That you don’t have to be employed and monetizing your labor to create anything worth sharing?

Even if you would be miserable without some form of external validation how can it be difficult to imagine that others are capable of simply enjoying the limited time they have to be alive, spending it on things that matter more to them than polishing a deck on SaaS sales metrics? How can it be unclear that the reality of work for the vast, vast majority of humans on earth is a brutally costly and undesirable experience? People die in low-paying jobs they took to survive roughly every other minute.

Good for you that you dig what you’re doing though, I hope you never have to see the alternative.

I think the issue is not that the LLM can’t optimize a straightforward algorithm but that what makes an algorithm successful in a game is much more nuanced than big O notation, especially where visuals are concerned.

Are they really? Nothing of value is produced unless the exact artifact is transmuted into a professional good? There’s nothing to be gained from sketching preliminary drafts or notes because they’re never polished for publication?

A few decades later, Venice brilliantly exploited the Fourth Crusade’s debt (and the fact that they built all the Crusade’s ships) to proceed on a debt-collection/revenge tour along the Adriatic, sacking Zara, getting the entire Crusade excommunicated, and ultimately diverting the entire force to Constantinople where they sacked the city in 1204 and replaced the government with the Latin Empire.

I found City of Fortune by Roger Crowley to be a very entertaining read, the full story of the Crusade is absurd. Venice completely dedicated its entire economy to producing the Crusade’s ships for years to make this happen.

I tried this too but it’s literally addressed in their docs already so I don’t think it’s much of a discovery

https://docs.typesafe.ai/model-jaggedness/jev-1.13#generatio...

Isn’t the value in being able to verify the design before implementing? It would be more ideal surely to prove correctness of the deployed code, but I still find it very useful.

I read this first and I thought the second example was random words you made up. Nope.

I just started working on some similar stuff, it’s been a dream forever but I never had time for the payoff to feel worthwhile. Would love to hear more about your Claude workflow and experience with Bevy, did you consider Godot?

Closing your laptop can also provide a clean, portable writing platform.

Individuals at risk may spiral into psychosis that is triggered or exacerbated by their use of AI. The term when used correctly is completely literal and in no way implies that AI itself is conscious.

The API has safety configuration for this

For most people, a description is better when it’s easier to understand and connect with. But you’re not most people.

NFT bacteria?

Really cool. Could see this being used for generative video game assets

How’d you find the sailing course? I’ve been planning to take a course on sailing and this sounds great.

Isn’t the selection bias a bit strong, though? I don’t think social media is a reliable source of conclusions about the community as a whole, or even the people creating the content in the first place.

Very clever that the user is prompted to use the autocomplete to read the description.

None of that seems like particularly nice behavior in retrospect. We could maybe also agree that the typical motivation behind taking things home was probably not to benevolently preserve historical records for future generations, although it’s nice that things worked out in this case.

They’re generally not very good at it, though. AI can be much more effectively tailored.

No, but it’s not clear that we’re likely to create sentience any other way

This seems to be focused on general intelligence. I always assumed it was more about the ability to learn skills and abilities rather than increasing IQ.

You’re right of course. The LLM is a calculator continuously predicting a best-fitting next token based on the data it was trained on.

If its outputs resemble human reasoning, it’s because the encoding and training process managed to capture those patterns and use them to simulate fitting text. There is no real reasoning happening or second-order thought, other than a simulation of that happening through the mimicry of human writing.

LLMs can’t be prompted to perform actual reasoning, but they can be told to generate “thoughts” about what they’re doing that bring out more nuanced detail when they give their answers. This isn’t any more magical than writing out a more thoughtful prompt to get a conditioned answer, it’s just getting the LLM to flesh out the prompt engineering for you in the general direction you want it to go.

That seems rather fundamental to me, the idea that with some generic prompting the model tries to fit what it thinks reasoning looks like and can then take advantage of the additional context that would others be buried too deep to influence its answer.

I suspect that prompting the model to explore “thought” asks it to go down paths of linguistic connections that are related to the topic but not immediately connected to the answer in a way that would immediately influence the top predictions. Bringing summaries of those connections into the token context is a kind of zero-shot training on their relevancy to forming an answer.

To me this is less “reasoning” and more suggestive of the idea that some of the heuristics for data retrieval and question answering we collectively refer to as reasoning have broader applications.

Just to be clear, you’re agreeing with the author that it’s primarily a moral distinction? Since the shortage is impacting all patients equally it seems harsh to support it.

It would help to understand why you feel that it’s immoral; my interpretation is that you are saying it gives people an advantage at a cost that you don’t want to be forced into paying yourself in order to compete.

Wasn’t aware that Gleam was sponsored by fly.io, but it’s nice to see the investment in the ecosystem along with their support for Phoenix. Excited to see more developments for Gleam!

This is the real issue with privatized AI imo. The limitations and guardrails exposed to the end user do not apply to the developers, who can alter results in ways that would be difficult to detect at best and invisibly manipulative at worst.

Don’t humans work similarly? Without an understanding of how to reason through a problem, as children we are likely to give the first intuitive answer that comes to mind. Later we learn techniques of logic and reasoning that allow us to break down a problem into components that can be reasoned about.

What this seems to show is that the model does not yet have system-level or architectural guidance on when to employ reasoning, but must be explicitly reminded.