How do you like them AIs
This week's main event was Apple Intelligence. Rather than recap the features, I want to showcase some additional news and takes.
Some additional perspectives:
Wired: Apple proved that AI is a feature, not a product.
Bloomberg: Apple is "paying" OpenAI for ChatGPT in exposure, not cash.
TechCrunch: Apple Intelligence is boring and practical - that's why it works.
The Verge: The AI upgrade cycle is here.
And of course, check out this week's deep dive if you haven't already.
Elsewhere in the FAANG free-for-all:
A new report details Amazon's struggles to build an upgraded Alexa, including bureaucratic woes and privacy concerns.
Google is bringing Gemini Nano to the Pixel 8 and 8a, plus ChromeOS is migrating to parts of the Android stack to bring AI features to market faster.
And Microsoft is delaying its new Recall feature after a backlash over potential security issues.
The ARC Prize
While many are concerned about the pace of AI progress, some believe we're not advancing fast enough. Thus, the ARC Prize - a new $1M competition intended to push AGI progress forward.
Between the lines:
The prize is funded by Mike Knoop (cofounder of Zapier) and François Chollet (creator of Keras).
The two believe AGI progress has stalled (and that OpenAI has set the industry back a decade), LLMs are a dead end, and new approaches (which ARC tests for) will be required.
The test relies on spatial reasoning and inferring rules from simple examples, reminiscent of Tim Lee's previous tests with leading LLMs.
Elsewhere in AI anxiety:
Perplexity was accused of plagiarizing news stories via its Pages feature, though the company says it has revenue-sharing deals in the works.
Opus Clip and HeyGen, two US-based startups, are scrambling to relocate their China-based engineers as AI geopolitics gets more tense.
And Tesla shareholders sued Elon Musk for starting a competing AI company.
Golden State of mind
California is often at the forefront of regulation, but a new statewide AI safety bill is getting a lot of pushback from Silicon Valley.
Why it matters:
The bill, which will be voted on by the general assembly in August, would create a new state agency to oversee AI development. AI companies would need to report to the agency and build "kill switches" into their models.
Almost every major LLM maker (OpenAI, Anthropic, Cohere, Deepmind, Meta) would be impacted, as they all operate within the state.
Critics suggest that the new legislation will push AI startups to leave the state and could stifle innovation and open-source AI development.
Elsewhere in AI regulation:
California isn't alone in regulating AI - in recent months, state lawmakers have proposed nearly 400 new laws on AI.
Other countries are jumping into AI with both feet - whether they're buying Nvidia chips to power sovereign AI companies or partnering with OpenAI to streamline their legal system.
And while Microsoft, Google, and Meta already have robust policy teams, new AI startups are playing catch-up: OpenAI has grown its global affairs team from 3 to 35 in the past year.
Speaking of OpenAI:
Elon Musk dropped his lawsuit against OpenAI and Sam Altman.
OpenAI hired Sarah Friar (formerly CFO of Square) and Kevin Weil (former VP of Product of Instagram) to their exec team.
And the company also appointed retired US Army General Paul M. Nakasone to its board.
Things happen
Clearview AI proposes settling class action suit with equity. Samsung cuts AI chip production time by 20% after new integrations. Generative AI takes robots a step closer to general purpose. Some future Microsoft AI products may switch from OpenAI's tech. Picsart teams up with Getty to take on Adobe’s ‘commercially-safe’ AI. The NYTimes covers "slop" - a new(ish) term to describe the internet's AI content. A look at the problems with building AI products. LinkedIn leans on AI to do the work of job hunting. Adobe clarifies its terms around customer content ownership and training. Google's AI system seemingly decreased stop-and-go traffic by 30% in test cities. Gemini and Copilot refuse to discuss elections, including whether Biden won in 2020. Mathematician Terence Tao on how academics can use AI. Helen Toner worries ‘not super functional’ Congress will flub AI policy. Yahoo resurrects Artifact inside a new AI-powered News app. AI detectors get it wrong - writers are being fired anyway. The rise and fall of BNN Breaking, an AI-generated news outlet. FTC Chair Lina Khan shares how the agency is looking at AI. "AI-powered" has become a red flag. Photographer disqualified from AI image contest after winning with real photo. Uncensor any LLM with abliteration. Teams of LLM agents can exploit zero-day vulnerabilities.