https://typefully.com/VJPajala_1 Building Intelligent Automation | Expanding expertise in AI | ChatGPT Ninja Wed, 18 Dec 2024 10:00:08 GMT Wed, 18 Dec 2024 10:00:08 GMT Typefully https://screenshots.typefully.com/screenshot?size=1200x640&url=https://typefully.com/VJPajala_1/card https://typefully.com/VJPajala_1/Os982SH New podcast was just founded: Google AI: Release Notes. There are only 2 episodes so far, but they are highly interesting, especially now that Google 're-entered' the AI game. A great companion to the Google DeepMind podcast, which has been running for a few seasons already. https://typefully.com/VJPajala_1/Os982SH Wed, 18 Dec 2024 10:00:08 GMT
There are only 2 episodes so far, but they are highly interesting, especially now that Google 're-entered' the AI game.

A great companion to the Google DeepMind podcast, which has been running for a few seasons already.



]]> Os982SH tweet false null https://typefully.com/VJPajala_1/1t23Kgp Sending LLMs to space? We have sent multiple messages into the cosmos for other civilizations to find. 🚀 Symbols of humanity & Earth's location. 🚀 Sounds and images of Earth. 🚀 Millions of pages of human knowledge. How about sending a talking LLM instead of static data? https://typefully.com/VJPajala_1/1t23Kgp Mon, 16 Dec 2024 10:00:13 GMT
We have sent multiple messages into the cosmos for other civilizations to find.

🚀 Symbols of humanity & Earth's location.

🚀 Sounds and images of Earth.

🚀 Millions of pages of human knowledge.

How about sending a talking LLM instead of static data?]]> 1t23Kgp tweet false null https://typefully.com/VJPajala_1/the-state-of-ai-w7TmOZq Ilya talking about the state of AI in recent presentation. Recommended to watch but a few takeaways: 🫧 Just scaling models is indeed platouing. We are running out of data - there is only 1 internet 🫧 The above does not mean AI progress is platouing. 🫧 Superintelligence is "Clearly what is built… https://typefully.com/VJPajala_1/the-state-of-ai-w7TmOZq Sat, 14 Dec 2024 21:01:44 GMT
Recommended to watch but a few takeaways:

🫧 Just scaling models is indeed platouing. We are running out of data - there is only 1 internet

🫧 The above does not mean AI progress is platouing.

🫧 Superintelligence is "Clearly what is built here"

🫧 Super intelligence will be and feel very different that what we have now. It will be agentic (current AI is not agentic in any meaningful way yet), cabable of reasoning, understands (doesn't get confused) and is self-aware.

🫧 True agentic behaviour will be highly unpredictable

https://youtu.be/YziJp-VmzFc?si=JbbIfbZra70l6wRk]]> w7TmOZq thread false null https://typefully.com/VJPajala_1/googles-quantum-leap-6J4opzx Google announced its new quantum chip Willow. 📍It made calculations in 5 minutes that would take our currently fastest classical supercomputers much longer than the age of our universe (10 septillion years). 📍It solved the 30-year-old bottleneck challenge of quantum chips becoming increasingly uns… https://typefully.com/VJPajala_1/googles-quantum-leap-6J4opzx Wed, 11 Dec 2024 10:00:21 GMT
📍It made calculations in 5 minutes that would take our currently fastest classical supercomputers much longer than the age of our universe (10 septillion years).



📍It solved the 30-year-old bottleneck challenge of quantum chips becoming increasingly unstable and error-prone during scaling. The new design actually reduces instability as you scale it up.

It's going to take some time before we get to practical implementations, but this was a major leap forward.

With applicable quantum computing, we could:

🌍 Perfectly optimize global supply chains, traffic systems, and energy grids, saving resources and reducing emissions.

🌍 Simulate complex molecules to discover new drugs and enable personalized medicine based on the individual's gene makeup.

🌍 Train massive AI models in hours or days (instead of months), accelerating research breakthroughs.

🌍 Model and optimize renewable energy transitions and predict climate changes with precision.

🌍 Develop quantum encryption methods that are immune to future attacks.]]> 6J4opzx thread false null https://typefully.com/VJPajala_1/5SicdAy I found a really cool AI tool called *Emergent Mind* that you can use as an AI assistant to keep up with the latest in AI research. There's so much happening all over the place nowadays that all assistance is humbly welcomed. 😅 https://typefully.com/VJPajala_1/5SicdAy Tue, 10 Dec 2024 10:00:06 GMT
]]> 5SicdAy tweet false null https://typefully.com/VJPajala_1/JC2WJHB Condensed Christmas calendar coming from OpenAI 👌 https://typefully.com/VJPajala_1/JC2WJHB Wed, 04 Dec 2024 19:05:31 GMT
]]> JC2WJHB tweet false null https://typefully.com/VJPajala_1/ySS6NPw Interesting 🤔. 4o just had thoughts. A glitch in the matrix or is it Orion? https://typefully.com/VJPajala_1/ySS6NPw Thu, 28 Nov 2024 10:00:08 GMT
]]> ySS6NPw tweet false null https://typefully.com/VJPajala_1/ORgujqT What if we could compress a century of progress into just a decade? https://typefully.com/VJPajala_1/ORgujqT Wed, 23 Oct 2024 09:00:04 GMT
]]> ORgujqT tweet false null https://typefully.com/VJPajala_1/4QrqbwW What stuck with me the most was the idea of the merging of artificial and organic intelligence, deviating from the usual narrative of opposition. What if we were to receive brain implants, and from the cloud, a few trillion neurons were added to our brain's neural network? https://typefully.com/VJPajala_1/4QrqbwW Thu, 03 Oct 2024 09:00:05 GMT

]]> 4QrqbwW tweet false null https://typefully.com/VJPajala_1/advancements-in-ai-for-complex-tasks-BUlQXvm This paper explores whether modern AI models like OpenAI’s o1 can handle complex tasks that require planning—crucial for advanced problem-solving. Previous large language models (LLMs), such as GPT-4o or Claude Sonnet 3.5, excel at language understanding but struggle with multi-step reasoning. Th… https://typefully.com/VJPajala_1/advancements-in-ai-for-complex-tasks-BUlQXvm Wed, 25 Sep 2024 20:22:26 GMT
Previous large language models (LLMs), such as GPT-4o or Claude Sonnet 3.5, excel at language understanding but struggle with multi-step reasoning.

The study shows that OpenAI's o1 represents a significant leap forward, solving nearly all tasks in the Blocksworld test with 97.8% accuracy, compared to 62.6% for LLaMA 3.1 405B (zero-shot) and 57.6% for Claude 3.5 Sonnet (one-shot).

In the more challenging Mystery Blocksworld, where tasks are obfuscated to make them harder to solve, o1-preview achieved 52.8% in zero-shot, far surpassing previous models like GPT-4o, which managed only 4.3% in one-shot, and LLaMA 3.1 405B, which managed only 0.8% in zero-shot.

These results indicate that o1-preview brings a substantial (10-40 fold) improvement in reasoning capabilities.

This was only just the preview version of o1 - we can only imagine the capabilities of the actual o1 model, let alone when it is multimodal with tools.

*Paper summary partially created in collaboration with GPT-4o]]> BUlQXvm thread false null https://typefully.com/VJPajala_1/the-future-of-ai-nyljm2n Sam Altman wrote yesterday about the juicy future of humanity and the potential arrival of super (artificial) intelligence within 'a few thousand days.' The blog contains many ideas that would warm the hearts of techno-optimists like myself, but for some reason, this is what stuck with me the most:… https://typefully.com/VJPajala_1/the-future-of-ai-nyljm2n Tue, 24 Sep 2024 19:01:12 GMT


The blog contains many ideas that would warm the hearts of techno-optimists like myself, but for some reason, this is what stuck with me the most:

'If we want to put AI into the hands of as many people as possible, we need to drive down the cost of compute and make it abundant (which requires lots of energy and chips).

If we don’t build enough infrastructure, AI will be a very limited resource that wars get fought over and that becomes mostly a tool for rich people.'

AI investments are currently at an insane scale globally, but from this perspective, that's exactly how it should be.]]> nyljm2n thread false null https://typefully.com/VJPajala_1/alD7R9L Midjourney + InsightFace 😅 https://typefully.com/VJPajala_1/alD7R9L Fri, 10 May 2024 10:36:25 GMT


]]> alD7R9L tweet false null https://typefully.com/VJPajala_1/mita-on-rag-vWv88qj Haluatko tietää, mitä tämä kuuluisa RAG oikein on? Kollegani Samuli kirjoitti aiheesta hienon blogitekstin 👍 "RAG (Retrieval Augmented Generation) on menetelmä, jossa kielimalliin yhdistetään tiedonhakujärjestelmä. (Jatkuu) Kun RAG-järjestelmän kielimallille esitetään kysymys, se hakee ensin tied… https://typefully.com/VJPajala_1/mita-on-rag-vWv88qj Tue, 23 Apr 2024 09:00:23 GMT
"RAG (Retrieval Augmented Generation) on menetelmä, jossa kielimalliin yhdistetään tiedonhakujärjestelmä. (Jatkuu)

Kun RAG-järjestelmän kielimallille esitetään kysymys, se hakee ensin tiedonhakujärjestelmän avulla tietoa määritellystä lähteestä ja generoi sitten vastauksensa löydettyjen tietojen pohjalta. (Jatkuu)

Näin kielimallin ei tarvitse luottaa ainoastaan parametriseen muistiinsa, vaan se voi hyödyntää suoraan tiedonhakujärjestelmästä saatavilla olevaa tietoa." (Jatkuu)

Kurkistus hypetetyn tekoälytermin konepellin alle – Miten RAG toimii ja mitä sen hyödyntäminen vaatii? (efima.com) https://www.efima.com/blogi/kurkistus-hypetetyn-tekoalytermin-konepellin-alle-miten-rag-toimii-ja-mita-sen-hyodyntaminen-vaatii]]> vWv88qj thread false null https://typefully.com/VJPajala_1/starting-ai-journey-5-steps-to-build-from-UIUx2Vs AI from scratch in 5 steps. If I would start exploring what AI is from zero right now, I would use these resources in this order: 🔹 Elements of AI – University of Helsinki: A great introduction to AI basics for beginners. 🔹 Building AI – University of Helsinki: Expands on fundamental concepts and… https://typefully.com/VJPajala_1/starting-ai-journey-5-steps-to-build-from-UIUx2Vs Mon, 22 Apr 2024 09:00:14 GMT
If I would start exploring what AI is from zero right now, I would use these resources in this order:

🔹 Elements of AI – University of Helsinki: A great introduction to AI basics for beginners.

🔹 Building AI – University of Helsinki: Expands on fundamental concepts and explores practical applications.

🔹 Practical Deep Learning for Coders – fast .ai: Focuses on practical, hands-on lessons to familiarize coders with AI techniques.

🔹 Machine Learning Specialization – Stanford University via Coursera: A comprehensive program by industry experts covering supervised learning, unsupervised learning, and best practices in machine learning projects.

🔹 Deep Learning Specialization – Coursera: Ideal for a deep understanding and hands-on experience with deep neural networks developed under Andrew Ng's guidance.]]> UIUx2Vs thread true null