rdedev

So I guess the new opus will not run on my drug discovery project. It's just a binary classifier to screen for new malaria drugs. Fable completely have up on that codebase citing bio security concerns. Seems like this domain will go unsupported by Antropic

My favorite is the history of chili pepper. It's something so fundamental to indian cuisine that I assumed chillies were native to India for a long time. His video shows the true history of how that spice travelled from the Americas to all over the globe

Putting things into perspective is not minimising the problem. We literally have to do this to prioritise where our efforts can be useful.

Your argument makes sense if ai datacenters were using something close to like alfafa farming but the difference between them is soo massive it does not make sense.

Reducing pollution is a much better problem to fight for

There is a thing called political power. The local baker may be the most vile racist out there but he cannot affect my life. The CEO bank rolling a political party can

The group includes Netflix, the most technically capable streaming company. It's sad that companies will only go down kicking and screaming even for the mildest of regulations

I'm pretty sure paris looked looked nothing like now 100 years ago. Some other buildings had to come down to put up that art noveau building. I'm pretty sure someone would have been complaining about how that changes the character of the city they grew up in.

To directly answer your question about where to draw the line a simple heuristic is to leave public buildings alone. This can even include a privately owned shop that is accessible to the public. They are part of the cities culture. Private homes in the other hand shouldn't be protected.

Reminded me of this quote: the problem with machine learning is that it's the machine that does the learning

I'm pretty sure the LLM will get fed up and start writing an RPC

Also > An API arbitration aberratically assigned at authorized access, abridged and annotated, analytically assuring absolute assurance

Cool that you wrote all the words starting with "a" but I don't understand what you mean

> It’s becoming cheaper and easier every day to start a company that will disrupt the established players and bring down prices.

The underlying assumption here is that there is always something established players won't try to buy out new companies or use their existing capital to screw over others

> The winners in the end will be consumers, and the losers will be the big AI companies

Right now that is not the case. Look at the PC industry. I worry for the autonomy of a consumer in the future. It's probably going to be something like here is your rental thin client PC with agents on a monthly rental plan. What's that? you want to build a game with your own gpu? No no. A consumer grade gpu does not make sense in this day and age. Just ask the agent to build your game. We need the gpu compute for better things

I guess this is the reason why the sunscreens from haruharu are now suddenly available in amazon

Due to "politics" my teams main responsibility went from developing ai agents to just testing out a chat it developed by another team.

When you spend 8 hours a day doing mind numbing tasks, tasks that won't help you land another job and is constantly under stress of being fired, its bad for your mental health

It's hard to imagine if you have not experienced it. The air would still be hot even after the sun sets in some parts of India. Usually when wind blows over you you feel cool. With hot air it's like a blow dryer in your face. Just thermal energy being dumped on you making you feel even worse

This is why I really like karapathy's idea of llms having spiky intelligence.

We would assume that if tasks A and B are closely related. Mastery in A would mean mastery in B but that doesn't always work with an LLM

I tried running fable on this ML model I've been building. It's basically a binary classifier to predict activity of a compound for a certain assay.

Fable detected that it's something to do with biochemistry and switched over to opus. Huh

The example I had was of Ramanujam. "It was revealed to me in a dream"

There is also the possibility that an LLM judge would be happy with some code that looks like LLM generated code. But a maintainer for a specific project might not merge it for stylistic reasons

> The main idea is we provide documents to the LLM and it asks lot of questions which clear ambiguity and possible misconceptions the LLM might have

This kind of works but the difficulty is that you have to be very explicit about everything. It was mentioned in a spec document that a particular excel file is treated as a source of truth throughout the whole company and it is treated as an append only database. The agent still decided to add a check to see if a previous row was modified. It pushed back on its decision when asked why it decided to do so. "What if someone entered it wrong and had to correct it"? Valid question but it's not my teams responsibility to check for it

This check makes sense from a traditional development view point and that's why the agent did it. I would say it's good practice too but it's beyond the scope of the project it was working on. If what you are doing is beyond the norm you have to watch out for things like this

During the time that this paper was written agents were not really a thing. I would be more concerned about centralisation of work itself as a bigger concern

I wish I could find it but Simon Clark, someone who specialized in climate science, had put out a video about how we were only recently able to model the AMOC and it's shifting patterns thanks to this measurement we were doing

Edit: probably not this one but atleast tells us why measurement is needed https://youtube.com/shorts/-X5EhUbzLTY?si=_N92PNUiTi3STat6

This reminded me of the geth from mass effect. They get smarter as more geth "agents" network together.

What if there is a worm that spread through thermostats and another that spread through smart fridges and they finally infect a laptop with a gpu. They can exchange notes while they are there. Fun times

This is true up until the point where someone sends you a crappy old version of a word document that breaks when you load it in Libre Office.

I had to install office after that

There is a wide chasm between writing code in python vs "write a star craft clone". And that is not where near writing python vs writing binary code.

To put in another way, we have been building abstractions to make things easier for us to code. With coding agents you don't even code in the first place. It almost feels like a logical fallacy to compare the two

How would you approach this problem if you are let's say token constrained due to per month limits set in your company?

What I've tried to do is make the bot write detailed spec documents, slowly building it over time as I explain the full problem.

It works for the most part but it's you have some non standard requirement, the agent seems to skip over that part of the spec document when it starts to code. Or it would have needless checks for situations that I said will never happen

You are missing the point of why AI is being hated so much. Sequencing was just a tool for you that made your job easier. Right now it almost feels like CEOs can't wait to use AI to fire everyone

Sites pay good money to appear on top search results. Looks like the future is going to be sponsored AI sources. It's going to be even more difficult to figure out if google is presenting you with actual information instead of just an ad

I would say that most improvements are in easily verifiable things like code or math. Atleast that's where all the amazing results seem to be coming from.

Other domains I am not sure but I've heard from people like Cal Newport that the rate of increase outside of code and math are not as equally impressive

It wouldn't be a problem if 1) people get a say in the matter and 2) those tax breaks don't require increasing taxes else where. So far with all the data centers built that does not seem to be the case.

IMO we are either limited by data or reaching the limits of what's possible with a transformer architecture. Hardware will get us efficiency but I am not sure if it will lead to smarter models

Right now there are state govts bending over backwards to provide cheap energy for data centers. The difference is being paid by people who live nearby through increased electricity costs. This is a tax with just extra steps

Is the underlying assumption that gitlab will see new paying users because more people are buying coding agents?

If gitlab thinks they are as famous as github i don't know what to say. They should have atleast positioned themselves as a better github alternative