Superficial Weekly AI Briefing #22

Have a tip? Email it to [email protected]
Have you been forwarded this email? Subscribe

Another week and more drama for OpenAI with accusations it based one of its new voices, Sky, on Scarlett Johansson’s character from every AI-nerds favourite movie, Her. OpenAI insisted it didn’t imitate Johansson’s character, but Johansson didn’t buy it, swiftly moving to sue the company. OpenAI CEO, Sam Altman, didn’t help matters by tweeting “her” after launch.

The relationship between AI and the creative industries is already strained and its safe to say this latest misstep hasn’t helped matters. To that end, Sag-Aftra, the US actors’ union, is backing the No Fakes Act, a bipartisan bill in the US Senate that seeks to protect performers from unauthorised digital replicas.

Whether you think OpenAI did or didn’t copy ScarJo’s voice, it will be interesting to see the case play out in court (if it gets there) and what implications this has for the wider industry.

^{Scroll down for last week’s poll result}

Quick Sync

You’re probably not going to get replaced by AI, but you will be replaced by someone who knows AI and does your job. Tech workers are racing to pick up AI skills as companies scramble for top-tier talent. WSJ

A comprehensive list of companies that have partnered with OpenAI including Reddit, NewsCorp + more. Fast Company

The AI Seoul Summit took place last week and resulted in the Seoul Declaration to promote the safe, innovative, and inclusive development of artificial intelligence. Specifically - the Declaration calls on leaders to align their AI strategies with international standards, focusing on safety, innovation, and inclusivity to enhance compliance, risk management, and ethical practices. Official Seoul Declaration

Google is now pushing seriously into rolling generative results directly into search, and results have been, well, mixed. A reminder that AI is best used to help us do our work - not to do our work itself. Pedestrian

Reuters reports that the US is planning another round of sanctions to restrict Chinese access to advanced AI software, building on existing measures targeting chips. Of course, much of AI is open source, and open-source models by their nature circumvent these restrictions, so it’s not clear how much impact these new sanctions will have in practice. Reuters

Europe's rules on artificial intelligence will enter into force next month after EU countries endorsed a political deal reached in December. For those of us who have been through the joys of GDPR, we get a repeat - companies outside the EU who use EU customer data in their AI platforms will need to comply, although not for awhile longer. Reuters

Opinion

Meta’s AI Chief, Yann LeCun, continues his attack on LLMs and why they won’t reach human level intelligence - in short - because of a limited understanding of logic, a lack of persistent memory, and because they don’t understand the physical world. LeCun - of course - is working to develop an entirely new generation of AI systems (very much separate from LLMs) based on an approach called “world modeling,” where the system builds an understanding of the world around it like humans do. Forbes

A nuanced and informative opinion piece by Tyler Austin Harper on the “big AI risk nobody is seeing” - in short, tech companies are racing to build AI products that replace what it is to be human. Another way of saying what we’ve always said - that AI is best used to augment how we work, not to do our work for us. The Atlantic, Superficial Blog

Elon Musk with another hot-take, predicting that jobs will become ‘kinda like a hobby’: ‘The AI and robots will provide any goods and services you want’. Unfortunately, Musk doesn’t specify when in the future his prediction will come true, so for those of us with pesky things called jobs, it’s not yet time to kick up our feet. Fortune

Industry Insights

64% of CEOs say that succeeding with generative AI will depend more on people's adoption than the technology itself and more than half of CEO respondents say they are pushing their organisation to adopt AI more quickly than some employees are comfortable with. More evidence enterprises should think as much about how they deploy AI as what AI they’re deploying, and make sure their teams are trained to use it. IBM

AI Adoption

Gannett, a US newspaper company that owns hundreds of titles, is using generative AI to add summaries to the top of articles. The Verge

Longer Reads

Anthropic has an interesting paper about getting inside the ‘black box’ of an LLM and working out some of the features and concepts it’s learnt. Anthropic

Deloitte’s ‘Women in AI’ report highlights the significant gender gap in AI (women hold only 26% of data and AI positions despite making up 47% of the overall labor force) and how his disparity limits the diversity of perspectives essential for effective AI development. Deloitte

^{Last Week’s Poll Result:
82% you said you are Highly Positive about AI’s impact on your industry, with only 3% sharing a Negative response.}