Dans le deuxième article, nous avons évalué la capacité d’AMIE à établir un diagnostic différentiel, seul ou en tant qu’assistant pour les cliniciens. Nous avons analysé 303 cas médicaux complexes issus du monde réel. AMIE a surpassé les cliniciens, et les cliniciens assistés par
RESEARCH
-

AI Matches Doctors in Simulated Text Consultations
By
–
Dans le premier article, AMIE a égalé ou surpassé des médecins généralistes lors d'une étude basée sur des consultations médicales simulées par messagerie texte avec des patients-acteurs validés. Bien que la méthodologie comporte d'importantes limites, les résultats montrent le
-

AI Crosses Turing Test Threshold: DeepMind’s AGI Plan Emerges
By
–
This Week in AI – A world on fast-forward
We’ve crossed a threshold. The Turing test is no longer something we aim for. It’s something AI quietly passes, without us noticing. DeepMind’s AGI plan may not be perfect, but it raises the right question: not how powerful will AI get -
Training Set Anomaly: Need to Write Update Post
By
–
Especially weird considering this one is part of the training set almost certainly and at scale. I should write an update post.
-

Climate Scientists Use MATLAB for Renewable Energy Innovation
By
–
From climate modeling to renewable energy innovation, technology is driving critical research to protect our planet. Explore how climate scientists and engineers are using @MATLAB to address sustainability challenges: https://
spr.ly/60120CeiM #EarthMonth #ClimateAction -
Grok 3 pricing and benchmark concerns versus Claude and Gemini
By
–
It’s not the cheapest model on the market. Grok 3 is on par with Claude 3.7 Sonnet, but pricier than Google’s Gemini 2.5 Pro. And benchmark results? Let’s just say xAI’s reporting has raised a few eyebrows.
-
Seed-Thinking-v1.5: Advanced AI Model Capabilities Showcase
By
–
Now, can you imagine how strong Seed-Thinking-v1.5 will be?
Seed-Thinking-v1.5: https://
github.com/ByteDance-Seed
/Seed-Thinking-v1.5/blob/main/seed-thinking-v1.5.pdf
…
DAPO: https://
arxiv.org/abs/2503.14476
VAPO: -

ByteDance DAPO VAPO Reasoning Models SOTA Algorithms
By
–
Recently, ByteDance created 2 RL algorithms for buliding reasoning models: DAPO and VAPO. According to the tech report of Seed-Thinking-v1.5, VAPO now stands as the SOTA solution in value-based methods, while DAPO establishes a new SOTA result for value-free approaches.
-

Huawei Pangu Ultra 135B Matches DeepSeek-R1 Performance
By
–
Huawei is gearing up to release Pangu Ultra—a 135B dense Transformer trained on 13.2T tokens with 8,192 Ascend NPUs. Despite fewer parameters, it matches DeepSeek-R1 in performance.
No full tech report or model yet. Check it out: https://
arxiv.org/abs/2504.07866 -

Balu Adsumilli Honored as 2025 IEEE Fellow Video Processing
By
–
Congratulations to Balu Adsumilli, who was inducted as a 2025 IEEE Fellow for his contributions to video processing and pervasive user-generated online video content! Learn about Balu's work at: https://
goo.gle/3XW5Eam
