Keshan Sodimana retweeted
Excited to launch Gemma 4: the best open models in the world for their respective sizes. Available in 4 sizes that can be fine-tuned for your specific task: 31B dense for great raw performance, 26B MoE for low latency, and effective 2B & 4B for edge device use - happy building!
Ideal scenario for a Sri Lanka win in this #T20WC26 : @OfficialSLC takes a heavy loss to Pakistan, Pakistan goes on to qualify for the final, ideally against India, allowing Sri Lanka to host both a semi final and the grand final. #T20WorldCup2026
330
Keshan Sodimana retweeted
Just announced new versions of Gemma 3 – the most capable model to run just one H100 GPU – can now run on just one *desktop* GPU!
Our Quantization-Aware Training (QAT) method drastically brings down memory use while maintaining high quality. Excited to make Gemma 3 even more accessible for more developers.
Keshan Sodimana retweeted
Did you know that over 140,000 participants tuned in for the Kaggle GenAI Intensive with @Google last year? The course is now back, taking place on Mar 31-Apr 4, it offers Google ML expert-led theory sessions, hands-on labs, vibrant community, a capstone challenge w/swag, and remains to be no-cost to all!
More info and registration: rsvp.withgoogle.com/events/g…

ALT An image promoting the Kaggle GenAI Intensive Course with Google. The event, taking place from March 31-April 4, 2025, invites participants to join expert-led theory sessions, hands-on labs, a vibrant community, and a capstone challenge with swag. The course is free of charge and open to all. Last year, over 140,000 participants joined the event.
Keshan Sodimana retweeted
Introducing Mistral Small 3.1.
Multimodal, Apache 2.0, outperforms Gemma 3 and GPT 4o-mini.
mistral.ai/news/mistral-smal…
Keshan Sodimana retweeted
Gemma 3 is a collection of lightweight, state-of-the-art open models built from the same research and technology that powers our Gemini 2.0 models. → goo.gle/43Ic5RV
Keshan Sodimana retweeted
Collective intelligence of ants solving a puzzle
nitter.gq/DamnThatsInter/status/…
Keshan Sodimana retweeted
it has been 2 hours since openai announced o3 and it has so far failed to solve
- Riemann hypothesis
- Quantum Gravity
- FTL (Faster Than Light travel)
- P=NP
- Grand Unified Theory
- Cure for cancer
- Gpt 4.5
clearly this shows ai has hit a wall and openai is about to go bankrupt
Keshan Sodimana retweeted
Congratulations to OpenAI on the release of o3. The results are impressive and it's important that this technology remains accessible to more than a few powerful companies. With hard work and determination I expect the open source community can catch up in 3-6 months. Let's do it
Keshan Sodimana retweeted
Did you know there was a 42-day bus service from Dover, England to Colombo, Ceylon?
The total distance was 22000km and the ticket price was £78 (~£2500 adjusted for inflation).
Established in 1958 by Penn Overland Tours.
Keshan Sodimana retweeted
That's because this dude doesn't know what education is.
He speaks of growing wheat, herding sheep, riding a horse, and so on, but in the era of these skills, this was the kind of education given to slaves.
Only a slave, a person who was owned as property, and used as a machine for a task, could be expected to do one task for his whole life.
A gentleman, or even a freeman of the lower classes, was not a machine for labor, but a person who could be expected to act in his own interests, and thus would need to do many different things throughout his life, depending on what served his goals at the time.
And he would need to be able to independently learn these tasks, rather than needing to be taught them in childhood.
Therefore if a boy was to formally educated, that might include some of gentleman's skills (riding, fighting with a sword, the management of finances), but his education was centered around what education really meant:
A fundamental grounding in how to live and thrive as an independent and free-willed person.
Thus, he was taught the seven liberal arts of classical antiquity:
- Arithmetic
- Geometry
- Music
- Astronomy
- Grammar
- Logic
- Rhetoric
These were not trade skills in the sense that they did not enable the performance of any particular trade or task, but that wasn't the point.
The point was that they taught the young gentleman how to think and learn.
By contrast, modern government schools were founded to train clerks and factory workers at public expense... a servant class with the specific skills necessary to be useful workers, but not the general education to be independent or question their betters?
Have you noticed which two of these arts are utterly absent from a modern government-school "education"?
That's right, logic and rhetoric. Logic is how to arrive at true conclusions from known facts. Rhetoric is how to persuade.
A servant educated in logic might notice that the things he is being told are false. A servant educated in rhetoric might notice the techniques that are being used to persuade him to act in the rulers' interests instead of his own.
If you conceive of your children's education as training in career skills, whether that be growing rice or programming a computer, you are preparing them to be slaves, not free men.
If you properly prepare them to be free men, what skills will be lucrative or useful twenty years from now is irrelevant, because they will be prepared to learn them.
In my opinion, the seven liberal arts of the modern world are:
- Logic: how to derive truth from known facts
- Statistics: how to understand the implications of data
- Rhetoric: how to persuade, and spot persuasion tactics
- Research: how to gather information on an unknown subject
- (Practical) Psychology: how to discern and understand the true motives of others
- Investment: how to manage and grow existing assets
- Agency: how to make decisions about what course to pursue, and proactively take action to pursue it.
Notice that you didn't learn any of these things in school, even if you went to a so-called "liberal arts" college. Instead, they taught you things about mitochondria and calculus and symbolism in Jon Steinbeck novels where a boy has a dog, and the dog dies.
That's because liberal arts, whether you define them as I have, or slightly differently, are the arts of the master, the arts that make one a master, and therefore not be taught in a school for slaves.
Worry less about which "career skills" AI will take over, and more about whether you are training to be, and training your kids to be, high-agency, perceptive, self-motivated people who can navigate an unknowable future with an adaptable mind.
Keshan Sodimana retweeted
These often-heard lines like, "consciousness is just a hallucination", "consciousness is just what it feels like to be an information-processing system", etc. sound nice, but they fail to address the hard problem, which is the existence of the subjective experience.
Hallucinations and feelings are things that happen to an agent that already has a subjective experience in the first place. The explanation is assuming the existence of the thing whose existence it was supposed to explain.
Keshan Sodimana retweeted
Gemini 1.5 Pro - A highly capable multimodal model with a 10M token context length
Today we are releasing the first demonstrations of the capabilities of the Gemini 1.5 series, with the Gemini 1.5 Pro model. One of the key differentiators of this model is its incredibly long context capabilities, supporting millions of tokens of multimodal input. The multimodal capabilities of the model means you can interact in sophisticated ways with entire books, very long document collections, codebases of hundreds of thousands of lines across hundreds of files, full movies, entire podcast series, and more.
Gemini 1.5 was built by an amazing team of people from @GoogleDeepMind, @GoogleResearch, and elsewhere at @Google. @OriolVinyals (my co-technical lead for the project) and I are incredibly proud of the whole team, and we’re so excited to be sharing this work and what long context and in-context learning can mean for you today!
There’s lots of material about this, some of which are linked to below.
Main blog post:
blog.google/technology/ai/go…
Technical report:
“Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context”
goo.gle/GeminiV1-5
Videos of interactions with the model that highlight its long context abilities:
Understanding the three.js codebase: youtube.com/watch?v=SSnsmqIj…
Analyzing a 45 minute Buster Keaton movie: youtube.com/watch?v=wa0MT8Ow…
Apollo 11 transcript interaction: youtube.com/watch?v=LHKL_210…
Starting today, we’re offering a limited preview of 1.5 Pro to developers and enterprise customers via AI Studio and Vertex AI. Read more about this on these blogs:
Google for Developers blog:
developers.googleblog.com/20…
Google Cloud blog:
cloud.google.com/blog/produc…
We’ll also introduce 1.5 Pro with a standard 128,000 token context window when the model is ready for a wider release. Coming soon, we plan to introduce pricing tiers that start at the standard 128,000 context window and scale up to 1 million tokens, as we improve the model.
Early testers can try the 1 million token context window at no cost during the testing period. We’re excited to see what developer’s creativity unlocks with a very long context window.
Let me walk you through the capabilities of the model and what I’m excited about!





























