There’s no shortage of things I think should change or improve about OpenAI, but as of now GPT-4 is still the most useful LLM.
@aibreakfast
-
Gemini Advanced falls short vs GPT-4 in usefulness and image handling
By
–
Gemini Advanced (Gemini Ultra 1.0) is nowhere near as useful as GPT-4. Just some first impressions: Cons:
-randomly refuses to interpret images
-says cannot interpret images “containing people”
-any controversial question involving politics gives the same canned response -
ChatGPT and DALLE comics top reader feedback in newsletter
By
–
ChatGPT + DALLE creating comics in each edition of the AI Breakfast newsletter started as an experiment 2 months ago, and now generates more reader feedback than anything I write about. Took a lot of prompt massaging to get the system down, but I like where it’s ended up.
-
AI and VR enabling global medical diagnosis via MedPaLM in two years
By
–
AI + VR for medicine.
— AI Breakfast (@AiBreakfast) 6 février 2024
Doctors could diagnose and develop treatment plans for patients around the world.
And by Doctor I mean Google's MedPaLM in ~2 years. https://t.co/eTuBiSgwspAI + VR for medicine. Doctors could diagnose and develop treatment plans for patients around the world. And by Doctor I mean Google's MedPaLM in ~2 years.
-
@aibreakfast — 2024-02-04
By
–
Ultra also crushes GPT-4V (vision) in 10 out of 10 benchmarks for image interpretation
-
Gemini Ultra surpasses GPT-4 on 7 benchmarks
By
–
Google’s Gemini Ultra was just confirmed for release on Wednesday. Ultra beats GPT-4 in 7 out of 8 benchmark tests, and is the first model to outperform human experts on MMLU (massive multitask language understanding).
-
Dolma: A 3 Trillion Token Open-Source Training Dataset for LLMs
By
–
A 3 Trillion (yes, trillion) token open-source training dataset for LLMs: Dolma is built from a mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. Now available on Huggingface if you have a spare 5.4 Terabytes
-
Integrating Brain-Computer Interfaces with LLMs for Real-Time Thought Editing
By
–
People underestimate the significance of this. Combine the Telepathy computer-brain interface with an LLM that can add to/edit your thoughts onto a device in real time.
-
Samsung Targets Fully Automated AI-Driven Chip Manufacturing by 2030
By
–
Samsung to build all-AI, no-human chip factories to rival NVIDIA. Their goal is to “remove all human workers from the entire chip manufacturing process by 2030.” If smart-sensor manufacturing gets completely automated in 5-6 years, then cheap US land with the most sunshine