doug_durham

The whole resource argument against AI is quickly becoming obsolete. Time to move on to a new argument.

Can you provide reasons other than "he's wrong"? There is no other technology that compresses so my knowledge into such a small space. A 24 GB diffusion model can generate a good approximation of any human relevant image from a description. Ask it to generate an image of the Tivoli fountain and it will do an impressive job. You can't use it as a map, but it gives an excellent representation.

Finally I don't think there are "dangerous ways" to think of any technology. This is just another tool.

Most of the discussions I've read about AI is as a tool. There is a narrative being created that its all about replacing humans. I've not seen that outside of a view influencers.

I don't think you can make the absolute statement that there is no audience or market for these tools. I enjoyed watching all of these videos. They are far from banal crap. They are deranged. They are not intentionally deranged, it is limitation of the tool. If a video artist were to employ these tools not to save effort, but instead to create an aesthetic it would be proper and good use. Just like non-AI video editors or color grading tools.

Handwriting is inherently ableist. A portion of the population is blessed with the inherent ability of fast, legible handwriting. A portion of the population is not. Typing is an equalizer that allows more people to participate in the conversation.

The same way that I verified all of the code that was written before AI. Just because you hand type the code doesn't mean that you don't have to test and verify. I can test much more thoroughly with AI that without.

Sorry but you are just wrong. Perhaps if you are working on tiny problems in very small projects. Beyond that AI is simply many times faster regardless of how fast you type.

You seem to have a narrow definition of "build it". LLMs don't have agency. If I build software using an LLM I built it. Just like if I build a house out of lumber from a lumber mill, I built it even though I didn't hand carve the 2x4s.

The llm is the better community tool. The important change is that you don't have to settle with someone developing a monolithic tool that happens to do what you need it to do. That was the way things used to need to be done because of the cost.

I don't agree that Anthropic's position is AI is all you need. My interpretation is that it's about developer acceleration. I've been following their messaging closely and it's not that you can fire all of your developers. It's mostly their developers saying they are somewhat more productive. They have published papers showing that productivity gains are uneven.

The point of cursive was make writing efficient with a quill pen. With a quill pen you run the risk of ink dropping off the tip if you lift it from the paper. Therefore cursive tries to keep the pen in contact with the paper most of the time. I've never thought of cursive as a speed tool. My printing has always been much faster than my cursive.

Unfixable errors? Why unfixable? Python is Turing complete. I can see difficult to fix, but not unfixable. LLMs lower the bar to refactoring code mistakes.

I don’t believe that “reasoning” is very useful in large code bases written by multiple developers. If you are trying to be axiomatic and prove to yourself that the code is correct, there are a thousand different ways the part of the code may not support your axioms in ways that are not apparent. It’s better to have a handful of invariants that are well communicated to the team and a lot of tests.

I’m sure in certain domains this makes sense. However, English is a poor language for doing reasoning in. More and more I’m relying on the code itself as the documentation. One of the superpowers of LLMs is reading code. and turning it into readable English. I don’t keep the English prose around. I delete it. In your example, I don’t see having two sets of artifacts. I see working with an LLM to generate a code base, which is the specification. You still need to have sets of requirements that list the invariants and other parameters. But the process becomes generating the code, and then having the LLM read the code to see if it meets the requirements and invariants.

Not my experience at all. There are a couple of linters to choose from, and uv is becoming the dominant packaging environment. In my experience it is much easier to write large maintainable projects in Python than in TypeScript. Python has more language features and certainly a much better ecosystem. Who knows where things will go. In the future language choice will be much less important. Languages are largely a human artifact.

I’m not referring to trivial functions nor classes. I talking about things like overloading operators, or changing the default idioms of the language. This means that a person can’t read the code without fully understanding the entire class structure. For the most part DSLs are highly personal. What one person thinks is a great formalism that make it easy to reason about the problem, to everyone else is spaghetti code.

Languages are never the answer. Lisp isn't superior to Visual Basic. They are both Turing complete. Humans can express good ideas in each. It's better to think of languages as tools. An impact drill is not superior to a simple rotary drill. You will have better luck building a deck with an impact drill, but both will do the job.

In commercial practice DSLs are an anti-pattern. Someone will create an under-documented DSL that only they understand and move on. The following programmers have to try to decipher this strange language. It is almost always a better idea to use standard language structures and features. Code is read more than it is written.

Nice rant with no evidence. I'll choose to ignore it.

My Hyundai Ioniq 5 has Car Play and knobs and buttons. It's not either or.

People engage with music in different ways. Some people focus on lyrics and the connection they make. Other people largely ignore lyrics and focus on the rhythms and patterns. Both are valid and both are very human. We do not need gatekeepers coming in and judging what is music and what is not music.

House music can bring me as much joy as listening to Bach performed by a skilled ensemble. It depends on where I am at mentally. Both are valid forms of human expression.

Your statement is so imprecise as to be meaningless. Is EDM made on an Abelton DAW human made? Even though the human didn't touch an instrument and used a robot drummer? What about a human who uses AI to generate snippets of music and then pastes them together in an emotionally compelling way, much like HipHop artists do for traditionally sampled music? AI is a tool. Low/no-effort work on the part of humans is the problem.

The NSA is government agency. They are certainly not training any world class LLMs. They probably have some specialized fine tunings of existing models, but that's it. They don't have the capacity.

There’s nothing new here. This is how open source software has been since its inception. It’s just the nature of reality.

As has been pointed out elsewhere, DMA isn't a privacy regulation. It is simply about competition. You can be in 100% compliance with DMA and poor privacy protections. This is the crux of the problem. How do you preserve the privacy of your customers while complying with regulations where the simplest path is to compromise your customer's privacy?

I have to lift up my hand and move my arm around to use it. With a trackpad all I need to do is move my hand over and flick my fingers for gestures. My wrist never moves.

Rust, Python, Java, Ruby, Scala, Swift to start with. These are languages with very wide adoption. Objective-C is very Smalltalk-like, but it is being phased out for Swift.

This is a truly important paper. It formalizes the intuition that many in the field have. We can stop wasting time doing formal analysis of LLMs. If you have a problem that requires formal verification, don't use an LLM. You can use an LLM to help you build such a system, but the LLM can't be the system.

You think that LISP and Smalltalk aren't widely used is because they weren't easy to implement in the late 1980's? There have been many languages that have risen to prominence in the 40 years since, yet LISP and Smalltalk remain niche languages.

I use a trackpad to avoid virtually all of the issues created by a mouse. The trackpad gestures in macOS are magical.