This Week in AI (4/22/23): AI music bans, GDPR woes, and Nvidia’s amazing new text-to-video

I combed through 500+ saved tabs on AI this past week to find the top items (below).

Because it’s hard to keep track of why something is important, I’ve added a sub point for each link to highlight its significance. Enjoy with your ☕!

News to Know (12 Key Developments)

AI-generated photo wins major photography award, but winner rejects prize [Link]

• The winner deliberately submitted an AI-generated piece to make a statement.

Nvidia unveils text-to-video model [Link]

• Please click the link to see it in action. It’s UNREAL and portends how crazy this year will be.

Compliance with GDPR will be difficult for ChatGPT, portending fines and ban [Link]

• Numerous legal experts think it will be near impossible for ChatGPT to fully comply with GDPR.

AI-Generated Song Mimicking Drake and The Weeknd Pulled from Streaming Services [Link 1], [Link 2]

• New details are still emerging here, actually! AI-generated music is raising lots of questions.

Reddit to start charging AI models for access to its archives [Link]

• AI models use large bodies of data, and content companies now want to cash in.

StackOverflow jumps on the API charge bandwagon as well [Link]

• StackOverflow’s extensive code examples were likely used to train OpenAI’s current models

Stability AI launches their own open-source language model, StableLM [Link]

• Best known for Stable Diffusion, they’re now moving to compete with ChatGPT

Google plans radical changes to their search engine [Link]

• Google races to play catchup, and the CEO swears they’re moving faster!

New Google DeepMind team formed out of two AI teams [Link]

• Two AI teams that formerly bickered are now one unit. Google’s survival is at stake here.

Michael Schumacher’s Family Threatens Suing German Tabloid Over AI-Generated Interview [Link]

• AI-generated content is at the center of numerous legal firestorms. This is just one of them.

Microsoft developers own AI chip as ChatGPT costs OpenAI an estimated $700k per day to run [Link]

• AI is expensive. ChatGPT is expensive. Microsoft is launching their own chip to cut costs.

Employees said Bard was “cringe-worthy,” but Google launched it anyways [Link]

• Wonder why Bard disappointed us at launch? It’s because Google didn’t listen to internal warnings.

Science Experiments and Things to Try

A beginner’s guide to autonomous agents [Link]

• What’s the hype around autonomous agents? 100k stars on GitHub makes this one of the fastest-growing software projects, ever. This writeup explains what it does and how you can play with it, right now.

MiniGPT-4 launched, runs on just 12GB memory, and can process images [Link]

• Multi-modal models can now run on personal computers. This one can process images like OpenAI’s GPT-4. Insane and a glimpse of the AI future.

Things you can do right now with AI that you no longer need to pay a marketer for [Link]

• Great though-joggers of how marketing is actively transforming now that AI is here. Good for any professional.

Meta open sources their animated drawings AI library [Link]

• Pretty fun to see in action [an a great example of the weird science coming out of the AI sector these days.

Notable New Research Papers this Week

• LLMs are learning to program with natural language [Link]

• Analysis of why ChatGPT falls short in comprehension [Link]

• Using LLMs to create data lakes [Link]

• Just 51.5% of LLM search engine responses fully supported by citations [Link]

• Gisting enables 26x compression of LLM prompts [Link]

Enjoy!