"GPT-5 probably sometime this summer." From latest OpenAI podcast:
LLMS
-
OpenAI Podcast: Sam Altman Discusses AGI and GPT-5 Future
By
–
Introducing the OpenAI Podcast—a series of conversations with the people shaping AI. @sama joins @andrewmayne on the first episode to talk about AGI, (wen) GPT-5, privacy, and what comes next.
-
Meta Launches Llama 4 with Mixture-of-Experts Architecture
By
–
Introducing "Building with Llama 4." This short course is created with @Meta @AIatMeta, and taught by @asangani7, Director of Partner Engineering for Meta’s AI team.
— Andrew Ng (@AndrewYNg) 18 juin 2025
Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its… pic.twitter.com/HxCND1Fs1TIntroducing "Building with Llama 4." This short course is created with @Meta @AIatMeta
, and taught by @asangani7
, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced the Mixture-of-Experts (MoE) architecture to its -
LLM limitations, tool usage, and a specific Gemini 2.5 Pro failure
By
–
I think it would be a worse puzzle. A good LLM knows its limitations and uses tools. (Raw LLMs also can’t compute SHA1, which is why I used it.) In another reply in this thread, Gemini 2.5 Pro is seen failing at this task because it thinks the fifth letter of “Madrid” is d.
-
ASI Scientific Foundations: Beyond LLMs and Learning Algorithms
By
–
part of an ongoing effort to think about what ASI could actually mean from a scientific perspective. > is superintelligence just some form of LLMs?
> what is the learning algorithm? > what is the data? read more here: -
Superposition Reasoning: Novel Approach to Chain of Continuous Thought
By
–
Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thought Zhu et al.: https://
arxiv.org/abs/2505.12514 #ArtificialIntelligence #DeepLearning #MachineLearning -
Fine-tuning data scarcity limits model improvement potential
By
–
It would if we had infinite fine-tuning data.
But we most definitely don't. -
Beyond Autoregressive: Rethinking Sequential Symbol Prediction in AI
By
–
This is a fundamental flaw of sequential (auto-regressive) symbol prediction.
The fix is simple: don't do auto-regressive symbol prediction. -
Continuous Embedding Space Superior for Reasoning: Theoretical Proof
By
–
It is intuitively obvious that reasoning in continuous embedding space is dramatically more powerful than reasoning in discrete token space.
This paper from @tydsh and team show that it is the case theoretically. -
II-Medical-8B-1706: Efficient Open Medical Model Outperforms Larger Competitors
By
–
Intelligent Internet introduced II-Medical-8B-1706, an updated version of its open medical model Capable of running on <8GB RAM, the AI outperformed Google’s MedGemma 27B across benchmarks despite 70% fewer parameters