20240131 A new way to build AI, openly ～TED Talk～

A new way to build AI, openly

I have watched the above TED Talk, and here is my summary and a opinion on it.

<summary>
He has been researching NLP for 19 years from master degree because he was attracted by this technology, which learn by itself and can categorize each word by only sentences. But he felt dangerous about recently closed LLM model. Key points he thinks are "transparency(：for accountability)", "values(：for how to consider what doesn't have a clear answer)", and "attribution(for whose right it is)". And quoting the open project Wikipedia, openness is also very important in AI
<opinion>
Open and close problems are often discussed all kind of area. I think we can't conclude which is better or not. But it is clear that human being have been piling up all kind of knowledge through history, and continuing now at status quo. I totally agreed with his three points being picked up in this talk, but to resolve them is not easy. We have to try and error, to discuss, to improve, continuously