- Starter AI
- Posts
- Oxford’s MedSAM-2, OpenAI face departures, Figure’s AI robot
Oxford’s MedSAM-2, OpenAI face departures, Figure’s AI robot
Changes keep coming.
Sthefania
August 07, 2024
Hello, Starters!
There's no doubt that one of the biggest players in the AI race is OpenAI; we don't need to remind you of that. However, being at the top comes with its challenges, and this year they've faced plenty.
Here’s what you’ll find today:
Oxford researchers present MedSAM-2
John Schulman departs OpenAI
Figure introduces its Figure 02 robot
OpenAI releases “Structured outputs in the API”
ByteDance launches an AI video generator
And more.
🩺 Oxford researchers present MedSAM-2 (2 min)
Meta's open approach with SAM 2 promised to spark innovation, and we're witnessing that with the introduction of MedSAM-2, which comes from the hands of University of Oxford researchers. This AI model can process entire series of medical images with enhanced segmentation in 2D and 3D.
MedSAM-2 achieves this by treating images like video sequences. Results across medical datasets reveal that it surpasses similar segmentation models. Also, a relevant aspect is that the model remembers reliable predictions through a "Confident Memory Back," which streamlines the process.
👋 John Schulman departs OpenAI (1 min)
At this point, OpenAI is familiar with making headlines. This year, although it has brought major announcements by the company, it has also got its sudden twists, and some of those are related to the departure of its members. John Schulman, one of the co-founders, recently announced that he's leaving the company to join Anthropic.
According to Schulman, the main reason behind his decision is that he wants to keep focusing on AI alignment, and he believes that Anthropic is a place that can offer new perspectives on these topics.
Moreover, Greg Brockman, president and co-founder of the company, is also taking time off from OpenAI. However, he has claimed that it is more of a "sabbatical" to "recharge."
🤖 Figure introduces its Figure 02 robot (1 min)
Figure, the robotics company, has presented its latest advancement in humanoids, Figure 02, a robot that harnesses the power of AI to potentially make a change in places like factories or even homes. This new version distinguishes itself from "Figure 01," launched last year, as it includes a speech recognition and output system based on custom AI models to enhance human interaction.
Both hardware and software have been updated in Figure 02. It includes six RGB cameras controlled by a Vision Language Model, a set of hands whose force is comparable to humans, and improved battery capacity.
🧑💻As a way of making developers' lives easier, OpenAI has released "Structured outputs in the API," a feature that reduces the reliance on other tools and increases interoperability between systems, as it assures that model-generated outputs align with JSON Schemas, even complex ones, an aspect most LLMs struggle with.
🎥The announcement of Sora prompted a surge of AI video generation tools, and ByteDance, the Chinese company behind TikTok, is also taking its step into the field with "Jimeng AI," a text-to-video model with several subscription plans that allow users to generate 168 videos per month.
⚡️Quick links
What did you think of today's newsletter? |