Jonjoe1001
Me and AI
I want to start by saying AI, or current-day LLMs, have a lot of concern and distrust around them, and I agree. The idea of these systems being smarter, faster, better and taking everything from us is a fear.
AI has come a long way since it was invented in 1956. Current frontier models have capabilities that give the people with them a lot of power. But I hope it doesn’t go that way.
Where I’m coming from
If you haven’t noticed, I have a particular interest in machine learning, evolution, and AI. I’ve always been fascinated by it and wanted to study it since I was young. I am currently pursuing a degree in machine learning engineering.
What I hope for
I hope for a world where AI can help us solve problems, like AlphaFold. Cures for diseases and new material science. And the ability to sort through and interact with massive piles of data, leading us as humans forwards towards a better world.
What worries me
I personally don’t know if this is the world the big tech companies are going for. We seem to be heading towards an AI sentience race, and just like human unpredictability, I fear that outcome could be the best thing we have ever done or the worst. That being said, I want to be able to mold it in the right direction.
Do I use AI?
Firstly, yes, I use AI. I also develop my own artificial intelligence (more on that later). For me personally, my frontier model of choice is Claude by Anthropic. I don’t use any other big AI company’s AI.
AI does these things for me: it automates my pipelines (aka git details and dev logs) and repetitive programming. I have used AI on programming projects such as a lot of the Limitless Lab, where personal projects I had done were fed to Claude to then match the format and put them on my website.
Using Claude day to day helps a lot for solo devs. It takes a lot of the repetitive stuff out of the work, so I can focus on the fun parts, the logic, and the meat and potatoes.
The AI I build
As promised, more on that! The AI I develop myself: Emergent Evolver, our AI drones, and DIANA and BLUE (DIANA’s predecessor).
To anyone who doesn’t trust AI
I agree with your fears, and I think we should be worried. But we need to understand machine learning is more than chatbots or weapons. It has helped, and can help, us in massive ways in science. A few examples:
- Proteins. AlphaFold predicted the 3D shapes of around 200 million proteins, a problem biologists had been stuck on for 50 years. Its creators shared the 2024 Nobel Prize in Chemistry.
- New antibiotics. In 2020, MIT researchers used a neural network to find halicin, an antibiotic that kills bacteria many of our current drugs can’t.
- New materials. In 2023, Google DeepMind’s GNoME found 2.2 million new crystal structures, around 380,000 of them predicted to be stable.
- Weather. GraphCast makes 10-day weather forecasts in under a minute, and beat the leading traditional forecast on most of the things it was tested on.
- Fusion energy. In 2022, a machine learning system learned to hold and shape the super-hot plasma inside a real fusion reactor in Switzerland.
Let’s talk
I’d love to hear what you think, whether you agree with me or not! Reach out on the contact page or come chat on Discord.