OpenAI's identity crisis
Everything you need to know in AI this week
Welcome back! Before we dive into this week’s top AI stories, a quick PSA for paying ChatGPT users: GPT-4 Vision finally rolled out to all ChatGPT Plus customers. To nobody’s surprise, I’ve spent…a lot…of time nerding out over the hundreds (!) of new things you can do with ChatGPT’s updates. Hit reply and let me know: Which update are you most excited about?
Checking in With OpenAI
Another week, another round of exciting news from OpenAI. Here are some of the biggest developments:
1) ChatGPT can access the internet (again). OpenAI officially launched an internet browsing feature via Bing. Now, Plus and Enterprise users will receive “current and authoritative information, complete with direct links to sources,” says OpenAI. Here’s why this is a big deal:
Previously, the chatbot was limited to data before September 2021.
In May, OpenAI tried to launch browsing—but quickly pulled the feature upon realizing that people used it to display paywalled content.
Since then, OpenAI updated the feature to follow robots.txt, the file that allows web crawlers to access a site's content. This protects paywalled content—but also means that some sites still may block ChatGPT’s access.
Regardless, bringing ChatGPT to 2023 is a milestone for the company. As OpenAI CEO Sam Altman said in a tweet, “we are so back.”
2) DALL-E 3 is now available in beta. A month after debuting the latest iteration of its image generator, users can finally access it directly within ChatGPT (just select “DALL-E 3 (Beta)” from the GPT-4 tab). The integration with the chatbot is making prompting DALL-E 3 way easier—ChatGPT refines your prompts for you.
Looking ahead: With OpenAI’s first developer conference coming up in November, we can expect more major updates soon. Anticipated updates are intended to reduce costs for developers using OpenAI’s technology, sources tell Reuters. Others are speculating that OpenAI might even introduce its own autonomous agent. 👀
But wait—there’s more: OpenAI quietly changed the “core values” listed on its website. The top value? A “focus on artificial general intelligence (AGI)”—AKA AI that doesn’t just match human intelligence, but surpasses it. The company’s new value statements also swap out words like “thoughtful” and “impact-driven” for “intense” and “scrappy.”
All this to say: Is OpenAI going through an identity crisis, or is something bigger brewing? ☕️
🗞️ Other AI News & Articles
Baidu Unveils Ernie 4.0
Of all the new chatbots on the block, very few actually have the capabilities to rival ChatGPT—but Chinese tech titan Baidu is up for the challenge. The company unveiled its new AI model, Ernie 4.0, at a conference in Beijing this week.
Ernie 4.0 is the latest version of the Ernie 3.0-Titan LLM, which powers Ernie Bot (China’s version of ChatGPT).
Baidu claims that Ernie 4.0’s sophisticated capabilities in reasoning, memory, and comprehension match those of GPT-4. (Since Ernie 4.0 isn’t publicly available yet, this bold claim has yet to be verified.)
Ernie 4.0 will be integrated into Baidu’s search, map, and cloud service products (similarly to how Google’s Bard integrates with Google Search, Maps, and Gmail).
But there’s one potential hurdle: With tightened U.S. trade restrictions on semiconductors, Chinese tech companies are in danger of being cut off from the advanced chips needed to create cutting-edge AI products. It’s still unclear how the new trade restrictions will impact Ernie 4.0’s widespread deployment; Baidu may have stockpiled American AI chips, but once their stash runs out, improving Ernie 4.0 could get way more expensive.
Ghostwriter’s Vision for the Music Industry
Remember when “Heart on My Sleeve,” the AI-generated song that cloned the voices of Drake and The Weeknd, broke the internet earlier this year? The ski mask-wearing creator of that song (AKA Ghostwriter) just proposed a potential solution for AI and the music industry to collaborate.
Ghostwriter proposed a clearinghouse approach: Artists license their voices to AI projects.
This ensures that artists get a cut of the revenue generated by AI songs and can set specific usage restrictions for their voice (e.g. no hate speech, no politics, etc.)
Zoom out: The inevitable intermeshing of art and AI is threatening to some, but it also poses an opportunity to democratize the music industry: With the right legal guidelines on AI-generated music, artists could increase earnings and exposure, while AI users can take their shot at making the charts. 🤝
But Ghostwriter’s proposal still hinges on the participation of big labels and artists—and according to Bloomberg, many are skeptical about this adoption.
Anthropic’s Claude is now available in 95 countries.
A 21-year-old student deciphered words from a burnt Roman scroll in an AI-enabled breakthrough.
Google pledged to defend users of its generative AI products against copyright claims.
Midjourney now has its own mobile app…sort of.
Researchers developed a new AI tool that can “future-proof” vaccines and prevent epidemics.
📺️ This Week’s Videos to Watch:
More important AI news: Dive deeper into this week’s hottest AI news stories (because yes, there are even more) in my latest YouTube video:
Exciting upgrades: Still wondering what you can actually do with the new GPT-4 Vision feature? I’ve got you covered:
More OpenAI: Still curious what to expect from OpenAI’s upcoming developer conference? This video digs into all the rumors and reports:
And there you have it! Thanks for all of your AI-themed Halloween costume ideas in response to Wednesday’s newsletter by the way—dare I say many of them were better than what ChatGPT came up with. Have a great weekend!
If you haven’t already, be sure to follow me over on X. I like to share news and AI updates as they happen over there. And don't forget to check out all of the newest tools we've just added on Future Tools!
You rock! See ya next week. :)
P.S. This newsletter is 100% written by a human. Okay, maybe 96%.