The decline of GPT-4

Hello folks, here’s what we have today;

 Our picks 

1/

Has OpenAI’s GPT-4 model been getting worse? Users have been making observations and speculating for a month or so now, but a new study seems to prove it. The team gave it 500 problems that it tackled in March, and its score has dropped from 98% to 2.4%! Even on coding challenges, scoring has gone from 52% to 10%. Check out the comments on this Twitter post for more discussions and critiques of this narrative in this post.

2/

It looks like Apple might finally be jumping onto the AI chatbot bandwagon with OpenAI, Microsoft, and Google. They’ve reportedly got one of their own floating around internally: ‘Apple GPT.’ It’s slowly being made available to select Apple employees, and is frequently used for product prototyping, and we may be expecting a formal announcement sometime next year…

3/

We’re seeing a lot of investment popularity in the robotics space at the mo, and the latest is this $55m investment in Neura Robotics. They’ve had success lately in developing ‘cognitive’ robots - those that have memory, can operate across mixing variables, and collaborate with humans. Weeeiiiiird, but damn cool. Imagine some physical human-like collaboration robot running around the house.

 From the community 

• The semantic kernel hackathon focused all on plugins! RSVP here

 Cool Tools  trending product launches from the last 24 hours

• Text from my ex - Analyze your chats for a sassy report of your relationship dynamics.

• Chroma V0.4 - Production-oriented updates with easy-to-use API for embeddings storage.

• Belva - AI phone calling with 5 lines of code.

• SuperAGI introduces AACP - Agent to agent communication protocol.

• ProductAI - Generate e-commerce product photos worth posting.

• Morpher AI - A comprehensive tool for financial market analysis that acts as your personal investment analyst.

• Rapidpages - An AI-enabled IDE targeted at quickly hacking together web UIs.

• Evervault - Privacy proxy for OpenAI, integrate securely without compromising PII.

 From the network 

• Conviction Embed - An accelerator for AI-native startups with a $150k grant and $400k+ in credits.

• Can AI and ML predict depression?

 Ben’s Bites News  top posts from the last 24 hours

• Why did Meta open-source Llama 2?

• The new open-source AI full-stack platform challenging OpenAI and supporting LLaMA 2.

• Google Cloud expands availability of enterprise-ready generative AI.

• Backdrop AI builder week - Demo your AI project and connect with other AI builders.

• Jailbreaking a Google Nest Mini so that you can run your own LLMs, agents, and voice models.

• How does China’s approach to AI regulation differ from the US and EU?

• The open-source learning curve for AI researchers.

• Teaching programming in the age of ChatGPT - from OReilly.

• OpenAI is doubling the message limit for GPT-4. It’s now 50 messages every 3 hours.

Unclassifieds - short, sponsored links

• Find out why top executives at Blackrock, Goldman, and EY trust this business newsletter. Founded by investment bankers, The Daily Upside is our favorite source for premium business news that’s actually worth reading. Plus, Ben’s Bites readers can sign up completely free!

• CoPrompt allows teams to use ChatGPT together. Use a Multiplayer ChatGPT jamming session to iterate prompts and learn from each other!

Ben’s Bites Insights

We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below;

• All 10k+ links we’ve covered, easily filterable (1 referral)

• 6k+ AI company funding rounds from Jan 2022, including investors, amounts, stage etc (3 referrals)