I’ll show the benchmarks in which PaLM (which now powers Bard) beats GPT 4, and detail how they use SmartGPT-like techniques to boost performance. Crazily enough, PaLM 2 beats even Google Translate, due in large part to the text it was trained on. We’ll talk coding in Bard, translation, MMLU, Big Bench, and much more.
I’ll end on the Universal Translator deepfakes and the underwhelming results from Sundar Pichai and Sam Altman’s trip to the White House and what Hinton says about it all. On a more positive note, I cover Med PaLM 2, which could genuinely save thousands of lives.
PaLM 2 Technical Report: https://ai.google/static/documents/palm2techreport.pdf
Release Notes Google Blog: https://blog.google/technology/ai/google-palm-2-ai-large-language-model/
Bard Access: https://bard.google.com/
Scaling Transformer to 1M tokens: https://arxiv.org/pdf/2304.11062.pdf
GPT 4 Technical Report: https://arxiv.org/pdf/2303.08774.pdf
Bard Languages: https://support.google.com/bard/answer/13575153?hl=en
Self Consistency Paper: https://arxiv.org/pdf/2203.11171.pdf
Are Emergent Abilities a Mirage: https://arxiv.org/pdf/2304.15004.pdf
Sparks of AGI Paper: https://arxiv.org/pdf/2303.12712.pdf
Big Bench Hard: https://github.com/suzgunmirac/BIG-Bench-Hard
Google Keynote: https://www.youtube.com/watch?v=cNfINi5CNbY
Gemini: https://www.youtube.com/watch?v=1UvUjTaJRz0
Med PaLM 2: https://www.youtube.com/watch?v=k_-Z_TkHMqA
TPU v5: https://ai.googleblog.com/2022/01/google-research-themes-from-2021-and.html
Hinton Warning: https://www.youtube.com/watch?v=FAbsoxQtUwM
White House Readout: https://www.whitehouse.gov/briefing-room/statements-releases/2023/05/04/readout-of-white-house-meeting-with-ceos-on-advancing-responsible-artificial-intelligence-innovation/