Did the technology which gave rise to AlphaGo directly lead to ChatGPT-style apps?

General Go Discussion
1

Can the lineage of chatGPT-style apps (such as Claude, Deepseek, Grok, Gemini) be traced back to the same technology which birthed AlphaGo in 2016? Or are these technologies independent of one another? The original Nature paper whereby Deepmind “released” alphago into the scientific community is far too advanced for someone of my dimwittedness… :persevering_face:

https://www.nature.com/articles/nature16961

My reason for asking is because it’s INCREDIBLY impressive that these tech companies were working on stuff that has profoundly changed the world in the last 1-2 years but they were able to foresee this back in the mid-2010’s? If so, that’s very impressive that tech companies foresaw the potential of neural networks. Or was the potential of neural networks speculative back then and (from a 2016 perspective) potentially a “money pit” where there might have been zero meaningful return on invested capital?

1 Like
2

Depends how directly you mean. They are both neural networks, but quite different architectures of them. The revolutionary paper which ushered in the world of LLMs was Attention Is All You Need - Wikipedia which came in 2017, after AlphaGo.

4 Likes
3

No. LLMs are mostly a different kind of neural nets. However, AlphaGo was directly relevant for Hassabis’s nobel price for protein folding research with AlphaFold2.

3 Likes
5

A neural net is an AI model that shovels calculations through multiple nodes repeatedly, instead of doing a single big calculation (this technique was originally based on an abstract guess about how brains work, hence the name).

Deep learning is an architecture for neural nets where there are many “layers” nodes. You can think of it as building neural nets on top of neural nets on top of neural nets.

Transformers are a twist on deep learning that made LLMs possible.

AlphaGo was a neural net, and a deep learning model, but not a transformer model. It may have been important to the dominance of deep learning inside Google’s research programs, but the Google Brain team that invented transformers was (at that time) distinct from the Google DeepMind team that built AlphaGo.