AI megathread

[Laird]

Google unveils its latest artificial intelligence model Gemini, claiming it largely outperforms OpenAI's ChatGPT (by AP/Reuters on 7 December, 2023)

The technology's problem-solving skills are being touted by Google as being especially adept in maths and physics, fuelling hopes among AI optimists that it may lead to scientific breakthroughs that improve life for humans.

In a virtual press conference, Google declined to share Gemini's parameter count — one but not the only measure of a model's complexity. A white paper released on Wednesday outlined the most capable version of Gemini outperforming GPT-4 on multiple-choice exams, primary school maths and other benchmarks, but acknowledged ongoing struggles in getting AI models to achieve higher-level reasoning skills.

Some computer scientists see limits in how much can be done with large language models, which work by repeatedly predicting the next word in a sentence and are prone to making up errors known as hallucinations.

"We made a ton of progress in what's called factuality with Gemini. So Gemini is our best model in that regard. But it's still, I would say, an unsolved research problem," Mr Collins said.

The inline video from the article is pretty impressive, featuring Gemini interpreting moving visual imagery:

Hands-on with Gemini: Interacting with multimodal AI