Weekly Top Picks #41
Code Llama / AI Napster moment / Llama 3 at ~GPT-4 level / Enterprises struggle to deploy AI products / ChatGPT finetuning / Google's TextFX / Gary Marcus on p(doom)
This week at a glance
Meta releases Code Llama: State-of-the-art (except GPT-4), open-source, 100K context.
AI is entering its ‘Napster’ phase and the hypocrisy of web scraping.
New rumor from Silicon Valley: Meta can already train Llama 3 (already ~GPT-4 level) and 4. They will open-source everything.
Generative AI inflated expectations: Companies struggle to deploy AI due to high costs and confusion.
OpenAI partners with Scale AI to allow companies to fine-tune GPT-3.5—and soon, GPT-4 too.
Google partners with rapper Lupe Fiasco to create TextFX: ‘AI-powered tools for rappers, writers and wordsmiths.’
Gary Marcus’ views on AI existential risk and p(doom) with an important twist and exemplary open-mindedness.