- Ben's Bites
- Posts
- Daily Digest: Gen AI in our daily life.
Daily Digest: Gen AI in our daily life.
PLUS: did NYT backstab OpenAI?
Sign up | Advertise | Ben’s Bites News
Daily Digest #321
Hello folks, here’s what we have today;
PICKS
OpenAI discusses its “love” for Journalism. The latest post from the AI company presents their position on training AI models on copyrighted content like news articles in response to the recent lawsuit from The New York Times.🍿Our Summary (also below)
Volkswagon is planning to add ChatGPT to its cars - VW wants to upgrade its current voice assistant, IDA. You could ask it stuff about your car or get directions just by talking to it casually. Roll out starts Q2 2024.🍿Our Summary (also below)
Duolingo cuts 10% of contractors as it uses more AI to create app content. Duolingo confirms the Reddit leaks (we reported) about this offboarding and says no full-time staff was impacted.🍿Our Summary
Rabbit Tech plans to unveil a mobile device with LLMs today at 10 am PT.
TOP TOOLS
Use offline AI on your iPhone with Mistral 7B.
Slack AI by EmbedChain - Build your own AI to chat with the unstructured data in your Slack.
EverArt - Your one-stop for AI curation, inspiration, and creation.
HN Jobs AI - Structured format for HN’s monthly hiring posts.
AutoWiki by Mutable AI - Generate a Wiki-style website to document your codebase.
Neuralhub - Design and build AI architectures.
Stylar - Your ultimate controllable AI image editor.
Spawn - Generate D&D character sheets with AI.
NEWS
Distribution is all you need.
The case for cyborgs - Augmenting human intelligence beyond AI.
A survey of 2778 AI researchers on timelines of AI advancement.
A simple guide to local LLM fine-tuning on a Mac with MLX.
AI trails:
Parag Agrawal, former Twitter CEO raises $30M from Khosla Ventures and others. LLM developers are the target audience, other company details are still unknown.
X's former head of trust and safety Ella Irwin joins Stability AI as senior VP of integrity.
Perplexity’s revenue has doubled to $6M ARR since October’s $3M reports.
Incumbent moves:
Samsung’s new smart home features include household maps with ‘AI characters’.
YouTube updates its policy to disallow AI content that simulates child victims and realistic violence.
Mayo Clinic pairs with Cerebras Systems to help develop AI for health care.
Getty Images and Nvidia bring generative AI to stock photos for SMBs.
QUICK BITES
OpenAI discusses its “love” for Journalism. The latest post from the AI company presents their position on training AI models on copyrighted content like news articles in response to a recent lawsuit from The New York Times.
What is going on here?
OpenAI believes training models on copyrighted content is fair use but provides their POV on the NYT lawsuit.
What does this mean?
OpenAI gives out 4 points to make their case:
Collaboration with news organizations: OpenAI has recently partnered with AP, Alex Springer, American Journalism Project and NYU. In return, they are helping reporters and editors with time-consuming tasks.
Training is fair use. OpenAI stands strong on this claim by linking to support comments and ruling in other countries.
“Regurgitation” is a rare bug. Exact memorization happens when the same content is in the dataset multiple times. OpenAI is working to solve this, and claims NYT’s small footprint in the training material doesn’t matter.
NYT is acting in bad faith. OpenAI and NYT were in talks before the court case and it claims that NYT didn’t inform OpenAI about this move. OpenAI makes its defense attack, stating that the examples shown by New York Times are cherry-picked, created by manipulated prompts with excerpts of those articles, and not an “allowed” use according to OpenAI’s terms anyway.
Why should I care?
Throughout, OpenAI pats its back with examples of their actions like linking to fetched articles, providing opt-out options, and removing the browsing feature in July.
Brilliant PR move from Open AI. The statement positions OpenAI as a “friend of journalists and NYT’s move as bad faith, anti-technology move. Will OpenAi be able to do so in court though?
QUICK BITES
Volkswagen will use ChatGPT to enhance its existing IDA in-car voice assistant, enabling more natural conversations between driver and car. The AI chatbot will be able to control vehicle functions, answer questions, and have more intuitive interactions.
What is going on here?
Volkswagen’s gonna start putting ChatGPT in its cars starting middle of 2024.
What does this mean?
VW wants to upgrade its current voice assistant, IDA. Right now IDA's pretty limited, but with ChatGPT it can have more natural conversations with drivers. So you could ask it stuff about your car or get directions just by talking to it casually. VW says it'll delete any data from the chats immediately too.
Still, ChatGPT sometimes makes stuff up when it doesn't know the real answer. So don't fully trust what it tells you! VW's working with a company called Cerence to try and make a solid car version.
Why should I care?
VW's the first company to officially say it's putting ChatGPT in cars. That's cool they're leading the pack. But I wonder if the AI will get confused about different car models and stuff like that.
Remember that Chevy dealer that used ChatGPT as a virtual assistant? People totally trolled it by asking to buy cars for $1 and writing Python for them. These AIs can go off the rails if you try to trick 'em.
But old-school car voice assistants are pretty stale and many other automakers are announcing their tests with generative AI. So, AI like ChatGPT will probably take over soon. VW's just getting a head start.
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)
Reply