We're probably in an AI bubble

Howdy, wizards.

The heated debate on legality and morality surrounding AI voices continues. OpenAI’s move to postpone their advanced voice mode seems to make more sense by the day, considering the lawsuits flying around.

Cloning your own voice, while admittedly a bit creepy, is at least not illegal – and it could help scale content creation. We’ve tested the best tools out there to do that, and found a clear winner.

‍

Dario’s Picks

The most important news stories in AI this week

‍

1. Venture capital firm Sequoia says we’re in an AI bubble, and they make a compelling case for it. The gist of the argument is that there’s a $600 billion gap – between revenue expectations and the numbers leading AI companies are actually racking up. Case in point, OpenAI, which has the biggest share of AI revenue is “only” at $3.4 billion.

Why it matters  The current lack of ROI is likely to negatively affect investors primarily. For startups and product builders, on the other hand, the outlook is likely less grim: the current investment boom is driving better products at lower costs, and a lot of learnings, which could improve their chances of long-term success.

PS this topic was just brilliantly covered by The Neuron – but it’s worth repeating here too.

‍

2. YouTube are trying to make deals with record labels to train AI music tools. YouTube is reportedly offering big, one-off payments to major record labels in exchange for licensing their songs, so they can train some new AI music tools on them.

Why it matters  YouTube is eager to build AI into its platform, but clearly taking a “permission first” approach. A smart move considering that the leading AI music generators, Suno and Udio, are currently being sued for having used copyrighted music to train their models.

‍

3. YouTube enables takedown requests for unauthorised AI-generated content. YouTube recently, quietly updated their policies that allows people to request take down of synthetic content resembling their face or voice.

Why it matters  YouTube is being proactive here, as this is likely to become a bigger challenge with rising accessibility and quality of voice cloning. Notable celebs have been vocal lately regarding unauthorised used of AI-generated content that mimic or resemble their voice; recent examples include Scarlett Johannson and Morgan Freeman.

‍

4. ElevenLabs’ new reader app can use voices of deceased celebrities. The company has partnered with estates of some iconic celebrities, including Judy Garland, James Dean, Burt Reynolds, and Sir Laurence Olivier. Their voices are available in ElevenLabs new reader app, which lets you listen to any digital text on your phone.

Why it matters  I don’t know about the morality of having your e-mail read back to you by “James Dean”. All I know is, unless some new regulation pops up soon, cloned voices will become ubiquitous. The ease of cloning and the practical utility of it is high, and people seem to be raving about reader apps like ElevenLabs and Speechify.

‍

5. Runway Gen-3 is now available to everyone – the most advanced video generator publicly available so far. To spark your creativity, Runway has released this neat guide on prompt structures, keywords and camera styles.

Why it matters  It’s amazing what the Gen-3 model can do (here’s a few examples). Keep in mind it can get pricey, fast: a 10-sec generation costs around 2$.

‍

Tool picks

The very best AI tools we’ve tried recently. We personally test and review AI tools to find the best one for every task. How we test & review tools.

‍

Recent top picks

 AI calendar  🥇 Reclaim AI

 Writing  🥇 Quillbot

‎ Meeting notes  🥇 Fathom

‎ Music Generation  🥇 Suno

‎ Text to speech  🥇Speechify

 Presentations  🥇Gamma‎

 Academic research  🥇SciSpace (40% off annual with WHATPLUGIN40)

‍

GPTs

Newcomers on whatplugin.ai’s top 1,000 list of GPTs in the last week. How rankings work.

‍