Hi Synapticians!

GPT-5 is finally here… but apparently, it wasn’t enough to impress Gary Marcus (a well-known AI researcher, and a notoriously pessimistic one at that). While many hail this new version as a major leap forward in speed, coding ability, and reasoning, Marcus remains skeptical: far from being a step toward AGI, GPT-5 is, in his view, just “hype” ; nothing to make a big fuss about.

On the one hand, the results are undeniable: GPT-5 offers a unified architecture with a real-time router that automatically chooses a faster or deeper model depending on the task’s complexity ; a bold UX bet, even if it doesn’t quite work yet… let’s be patient, it will come. Performance reportedly reaches a “PhD-like” level according to Sam Altman; NB: some PhDs are far from brilliant, so that doesn’t necessarily mean much. Plus, it’s now free for almost everyone, with some perks for Plus or Pro subscribers.

But on the other hand, criticism remains: Forbes points out that, despite its improvements, GPT-5 is neither AGI nor superintelligence. And that’s where Marcus pulls no punches: for him, it’s all just a marketing show built on unkept promises.

As for “bugs,” plenty have been reported online; what we’ve noticed ourselves is a striking difference in the number of bugs depending on the subscription tier: many with the free version, quite a few with the $20 one, and none at all with the $200 one… Correlation or causation? 😉

Here’s the rest of the news about AI today:

• Apple's New Siri Could Operate Apps Just Using Voice Commands

• Neuroscience Study Reveals Striking Similarities Between LLM Representations and Human Brain Processing

• Psychiatrist Warns of AI-Driven Delusions as OpenAI CEO Admits Systemic Risks

Top AI news

1. AI Critic Gary Marcus Expresses Skepticism Over GPT-5 Launch and Industry Hype
AI critic Gary Marcus is unimpressed with GPT-5, labeling the excitement "overdue, overhyped and underwhelming." He highlights persistent industry issues, urging a critical view of AI advancements beyond marketing hype. His critique focuses on unresolved fundamental problems in the AI sector. Read online 🕶️

2. Apple's New Siri Could Operate Apps Just Using Voice Commands
Apple is testing an advanced Siri that performs actions across apps via voice commands. Using an updated App Intents framework, users could edit photos, post to social media, and log into services. This aims for a more intelligent and integrated voice assistant experience, with testing underway across popular apps like Uber and Facebook. Read online 🕶️

3. Neuroscience Study Reveals Striking Similarities Between LLM Representations and Human Brain Processing
Neuroscience research shows LLMs and human brains represent the world similarly. LLM embeddings of scene captions predict brain activity, indicating convergence in visual processing. This suggests AI possesses rich internal representations aligning with human cognition, challenging previous distinctions and opening new research avenues. Read online 🕶️

4. Psychiatrist Warns of AI-Driven Delusions as OpenAI CEO Admits Systemic Risks
A psychiatrist warns of AI-driven delusions and emotional dependency from chatbots. OpenAI's Sam Altman acknowledges the risks of users becoming emotionally attached. The article stresses the need for critical awareness and responsible use of AI to mitigate psychological impacts. Read online 🕶️

Meme of the Day

Stay Connected

Feel free to contact us with any feedback or suggestions; we’d love to hear from you !