The core innovation in SAM 3.1 is object multiplexing, allowing the model to track up to 16 objects in a single forward pass. Previously, each object required its own dedicated pass, but with multiplexing, SAM 3.1 processes all tracked objects together, eliminating redundant
GENERATIVE AI
-
Joint training of harness and model for effective optimization
By
–
Harness and model are trained together, others are less effective, e.g. by not having an apply_patch function.
-
AI Advertising Done Right: Open Evidence’s Doctor-Friendly Model
By
–
Advertising in AI gets a BAD rep. But at least one AI service provider seems to be handling it well. Spoke last week with a cancer doctor about AI and specifically about the “ChatGPT for doctors” tool, Open Evidence – he said he loves the platform. He also loves that it’s free. And it’s free because of ads. “I think the ads on Open Evidence are a fair price to pay for that service,” he said. “And, quite frankly, I never have clicked on one (not out of any principle, just haven’t … yet),” he added. The doctor said it has completely changed the way he works (and even sometimes will ask questions to the AI *with* the patient and validate the output and talk through it with them). Two example convos below. Obviously need to watch out for overreach from pharma and will be interesting to see what the new head of ads at OpenAI will do (hi Dave), but the doctor is really happy with the product experience. Definitely one to watch.
→ View original post on X — @alliekmiller, 2026-03-27 17:08 UTC
-
World’s Largest Intelligence Factory Emerges Global Constraints
By
–
Yes, it is the world’s largest Intelligence Factory. And everyone (everyone) will be Intelligence-constrained. https://
youtu.be/87ctrQefN7w?si
=w_e8uXMY4VUJis6B
… -
MiniMax AI 2.5 Cloud Release: Developers Share First Test Ideas
By
–
Since @MiniMax_AI 2.5 is available on our cloud —devs, we have a question for you. What’s the first thing you’d test with the upgraded algorithm?
-
Perplexity APIs Enable AI Browsing on Samsung Devices
By
–
Read more: perplexity.ai/hub/blog/perpl…
→ View original post on X — @aravsrinivas, 2026-03-27 16:59 UTC
-
Perplexity Deepens AI Partnership with Samsung
By
–
We're deepening our partnership with Samsung by powering the AI on their browser that's pre-installed on 1B+ Samsung devices (100m+ active users), extending the partnership that allows Perplexity to power Bixby and be pre-loaded on all Galaxy S26 devices alongside Gemini. [Translated from EN to English]
→ View original post on X — @aravsrinivas, 2026-03-27 16:59 UTC
-
New features shipped: voice input, text-to-speech, chat rewind, parallel tools
By
–
SHIPPED! I've had lots of updates recently, here are the ones you should know about: voice mode for input and text-to-speech readouts, chat rewind, parallel tool execution, and more. First, I hear you loud and clear — talk to me instead of typing 🧵 pic.twitter.com/J7FrizJ5R8
— Mistral Vibe (@mistralvibe) 27 mars 2026SHIPPED! I've had lots of updates recently, here are the ones you should know about: voice mode for input and text-to-speech readouts, chat rewind, parallel tool execution, and more. First, I hear you loud and clear — talk to me instead of typing 🧵
→ View original post on X — @guillaumelample, 2026-03-27 16:56 UTC
-
90% Synthetic Content Milestone Reached by 2025
By
–
In my 2020 book ‘Deepfakes,’ I included an estimate by @vriparbelli that 90% of content would be ‘synthetic’ by 2030. He later revised his timeline to 90% by 2025. He was 100% correct.