Immanuel Giulea (@immanuelg) 's Twitter Profile
Immanuel Giulea

@immanuelg

#AI and #robotics

ID: 43782326

calendar_today31-05-2009 23:47:49

27,27K Tweet

4,4K Followers

4,4K Following

Logan Kilpatrick (@officiallogank) 's Twitter Profile Photo

Introducing the File Search Tool in the Gemini API, our hosted RAG solution with free storage and free query time embeddings 💾 We are super excited about this new approach and think it will dramatically simplify the path to context aware AI systems, more details in 🧵

Haider. (@slow_developer) 's Twitter Profile Photo

what i'm looking for in Gemini-3 - intelligent agents: work like personal AI assistants - logical reasoning: think through problems, not just follow scripts - multi-modal integration: seamless image, audio, video - long context window: handle "millions of tokens" without

Brett Adcock (@adcock_brett) 's Twitter Profile Photo

One of my proudest moments was sending an autonomous humanoid robot into the real world To this day, I believe Figure is the first and maybe the only company to achieve this

Derya Unutmaz, MD (@deryatr_) 's Twitter Profile Photo

Curing all diseases by the end of the century? That’s way too late and pessimistic! The target date in the age of AI should be no later than 2040!

Mark Kretschmann (@mark_k) 's Twitter Profile Photo

Google may have had a breakthrough with Continual Learning AI models: "We introduce Nested Learning, a new approach to machine learning that views models as a set of smaller, nested optimization problems, each with its own internal workflow, in order to mitigate or even

Google may have had a breakthrough with Continual Learning AI models:

"We introduce Nested Learning, a new approach to machine learning that views models as a set of smaller, nested optimization problems, each with its own internal workflow, in order to mitigate or even
Google AI Developers (@googleaidevs) 's Twitter Profile Photo

Gemini 2.5 Pro’s multimodal reasoning is delivering 2x more accuracy for complex financial document intelligence 📈 Pascal AI used it to build their knowledge graph, accurately parsing complex charts and tables where other models hallucinated.

Gemini 2.5 Pro’s multimodal reasoning is delivering 2x more accuracy for complex financial document intelligence 📈 <a href="/pascalailabs/">Pascal AI</a> used it to build their knowledge graph, accurately parsing complex charts and tables where other models hallucinated.
Chubby♨️ (@kimmonismus) 's Twitter Profile Photo

Some more big leaks: -GPT-5.1 -GPT-5.1 Pro -GPT-5.1 reasoning The release can't be far off, probably to compete the Gemini 3.0 pro release. h/t Lisan al Gaib for finding it first!

Some more big leaks:

-GPT-5.1
-GPT-5.1 Pro
-GPT-5.1 reasoning

The release can't be far off, probably to compete the Gemini 3.0 pro release.

h/t <a href="/scaling01/">Lisan al Gaib</a> for finding it first!
Rohan Paul (@rohanpaul_ai) 's Twitter Profile Photo

⚙️ Google Research introduces Nested Learning which stacks learning levels with different update speeds to cut catastrophic forgetting, and showcases a model called Hope as the first example. Treats the network design and the optimizer as the same thing viewed at different

⚙️ Google Research introduces Nested Learning

which stacks learning levels with different update speeds to cut catastrophic forgetting, and showcases a model called Hope as the first example.

Treats the network design and the optimizer as the same thing viewed at different
Peter H. Diamandis, MD (@peterdiamandis) 's Twitter Profile Photo

I think it's fair to say every child playing video games today already understands digital assets better than most CEOs. The next generation gets it...

Jubal Hardin (@jubal_hardin) 's Twitter Profile Photo

The cures can be ready by 2040. It will take until the end of the century to overcome the roadblocks set up by major pharmaceutical companies to allow those cures to get regulatory approval.

Bindu Reddy (@bindureddy) 's Twitter Profile Photo

I can literally feel OpenAI's compute constraints when I use GPT-5 Sometimes it's so unbearably slow that it appears to be just waiting for a spare GPU It's only natural for them to want to spend $1.5T Interestingly, neither Anthropic nor Gemini is anywhere near as slow.

Immanuel Giulea (@immanuelg) 's Twitter Profile Photo

New paper published in Nature concluded that : using widely used benchmarks for evaluation, the maximum capability density of open-source LLMs doubles approximately every 3.5 months. nature.com/articles/s4225…