This week, the TechCrunch crew (including yours truly) is at TC's annual Disrupt conference in San Francisco. We have a packed lineup of speakers from the AI industry, academia, and policy, so in lieu of my usual op-ed I thought I'd preview some of the great content headed your way.

My colleague Devin Coldewey will interview onstage the CEO of Perplexity, Aravind Srinivas. The AI search engine is flying high, having recently crossed 100 million queries served per week; on the other hand, News Corp's Dow Jones filed a lawsuit against the firm regarding what the publisher labels "content kleptocracy".

In a fireside, TC's transportation editor Kirsten Korosec will sit down with Zoox co-founder and CTO Jesse Levinson, who has spent a decade deep in the guts of autonomous car technology. And is now getting the Amazon-owned robotaxi company ready for its next big adventure-and we'll be reporting on it.

We will also hear from a panel on How AI Is flooding the web with disinformation—by Meta Oversight Board member Pamela San Martin, Center for Countering Digital Hate CEO Imran Ahmed, and UC Berkeley CITRIS Policy Lab founder Brandie Nonnecke. As more consumers gain access to the broader generative AI toolkit, the trio says the current crop of bad actors using these tools range from sophisticated state actors making deepfakes to disinformation in all its varied flavors.

And we'll hear from Cara CEO Jingna Zhang, AI Now Institute co-executive director Sarah Myers West, and ElevenLabs' Aleksandra Pedraszewska on AI's legal and ethical minefields. AI's meteoric rise has created new ethical dilemmas and exacerbated old ones, while lawsuits drop left and right. This threatens both new and established AI companies, and the creators and workers whose labor feeds the models. The panel will tackle all of this-and more.

That's just a sampling of what's on deck for this week. Expect to see appearances from AI leaders like U.S. AI Safety Institute director Elizabeth Kelly, California senator Scott Wiener, Berkeley AI policy hub co-director Jessica Newman, Luma AI CEO Amit Jain, Suno CEO Mikey Shulman, and Splice CEO Kakul Srivastava.
News
The Launch of Apple Intelligence: iPhone, iPad, and Mac users can start experiencing the first batch of AI-driven Apple Intelligence capabilities from Apple with a free software update.

Bret Taylor's new startup raises new money: Sierra, the AI startup co-founded by OpenAI chairman Bret Taylor, raised $175 million in funding round that values the startup at $4.5 billion.

Google expands AI Summaries: Google Search's AI Summaries, which give users a preview of what they are looking for at the top of the page of results, have started to roll out in more than 100 countries and regions.

Generative AI, e-waste: The new generation of computer appetite will see the industry produce enough 'e-waste' to equal more than 10bn iPhones being discarded each year, the researchers claim.

Open source, defined at last: The venerable institution dedicated to defining and "stewarding" all things open source, the Open Source Initiative, this week released version 1.0 of its definition of open source AI.
Meta releases its own podcast generator: Meta has published an "open" implementation of the viral generate-a-podcast feature in Google's NotebookLM.

Hallucinated transcriptions: Researchers claim OpenAI's Whisper transcription tool has hallucination problems. Whisper has reportedly admitted everything from racial commentary to imagined treatments into transcripts.

Research paper of the week
Google claims it trained a model to turn photos of handwriting into "digital ink".

The model, named InkSight, is trained to recognize written words on a page and then output strokes roughly resembling handwriting. According to the researchers behind the project at Google, this was meant to "capture the stroke-level trajectory details of handwriting so that a user can store the resulting strokes in the note-taking app of their choice.".

But isn't perfect. Google acknowledges that it sometimes gets it wrong. But claims that the model does better than average in a set of scenarios, including extreme lighting conditions.

Here's hoping it's not used to forge signatures.

This week's model: InkSight
Cohere for AI, the research lab from the nonprofit set up by AI startup Cohere, announced a new generation of text-generating models called Aya Expanse. The models can write as well as understand text in 23 languages and were said to outperform certain benchmarks with models such as Meta's Llama 3.1 70B.

According to Cohere, its secret to training Aya Expanse was a technique it has termed "data arbitrage." In drawing inspiration from the way a human learns new special skills by finding different teachers who can teach those specific areas, Cohere picked the multilingual "teacher" models able enough to generate synthetic training data for Aya Expanse.

Synthetic data isn't perfect. One set of research warns that overreliance results in models whose quality and diversity deteriorate worse each generation. But Cohere counters that data arbitrage does a great job of mitigating that effect. We'll soon know if the company is justified in its boast.

 Grab bag
As announced, OpenAI's realistic-sounding voice feature called Advanced Voice Mode for its ChatGPT is free in the ChatGPT mobile app for all users in the EU, Switzerland, Iceland, Norway, and Liechtenstein. Formerly, users had to pay for the ChatGPT Plus subscription to use the Advanced Voice Mode.

A recent piece in The New York Times highlighted the upsides and downsides of Advanced Voice Mode: relying on tropes and stereotypes in trying to communicate in ways that users ask. Advanced Voice Mode has gone wild on TikTok for its uncanny ability to mimic voices and accents. But some experts fear it could lead people to emotional dependency on a system that has no kind of intelligence — or empathy for that matter.