The week in AI: text-to-music, secrecy, and job replacements

Every week is a big one in AI and there’s often too much to read – we highlight eight of the big AI stories from the week that we think matter to the business of effective marketing. 

The internet is becoming less and less personal: social media is increasingly populated by professional creators; bots and spam are on the rise; and pivots to AI at Klarna and Apple raise big questions. The week in AI has seen the spotlight fall on the giants while smaller-scale systems change the texture of the online experience. 

News from the giants

Google Gemini image generation controversy shows the problem of guardrails after the search giant’s AI app made up some racially problematic results. It led some US conservatives to read anti-white bias into the company’s system.

• As observers have pointed out, it’s not a fundamental flaw in the technology as much as it’s an issue with the guardrails set up to correct biases in training data. Moreover, it highlights the risks of opening up creative tools while maintaining trust among its vast user base. 

What NVIDIA does and why it matters following last week’s eye-popping financial results (profits up +700% YOY) it’s worth asking what this extremely B2B company actually does.

• Thankfully, Oliver Feldwick, head of innovation at The&Partnership and regular WARC contributor, has written an explainer of what the company does, why it matters, and where it’s going. It has been shared many times around the WARC editorial department. 

Microsoft diversifies AI interests by sealing a new deal with the French artificial intelligence startup Mistral, which is developing open source models in contrast to OpenAI’s GPT closed models: a hint that Microsoft sees AI regulation shifting to encourage an ecosystem beyond the data centre’s operating powers.

Apple’s AI secrecy has investors looking for details – for a company that locks down leaks and speaks only when it is ready to do so, investors are now keen to hear about the company’s plans in the field of artificial intelligence following its sunsetting of its electric vehicle project.

• It’s possible that shareholders may vote to compel the company to report on its AI operations and to disclose ethical guidelines. What Apple is saying is that it plans to “break new ground” with AI this year.

Klarna talks up AI for customer service – the buy now, pay later fintech is gearing up for an expected IPO, with aggressive expansion. It is now touting its OpenAI-powered customer service assistant that it says is doing the work of 700 people (following heavy layoffs). 

New tools

• Text to video approaches reality beyond Open AI as Chinese state TV sees the first AI-developed cartoon series, based on a training dataset of China Media Group’s archive, according to the channel.
• Adobe debuts a text-to-music tool in the form of Project Music Gen AI Control, which generates a basic (and then easily adaptable) basic tune from text descriptions. While still in the research stage, the technology speaks to another area of advertising production set to be altered by AI.
• The BBC updates AI plans in a short blog post that is worth reading. It explores how the technology’s real capabilities go far beyond the creation of content – such as cross-format repurposing, translation, headline help, and streamlined indexing – all of which point to much richer possibilities than AI-written articles. What this means for staff headcounts, meanwhile, remains to be seen.

News from the spam tsunami

Bots and generated images are proliferating, especially across the increasingly wild social media landscape on X (FKA Twitter) as academics see bot traffic soar. Elsewhere, ghost kitchens – which sell exclusively through delivery apps – have witnessed not just enhanced images of food but entirely fabricated images of food, sometimes physically impossible items.

Sourced from WARC, Semafor, Financial Times, LinkedIn, Fast Company, SCMP, TechCrunch, BBC, ABC, 404 Media. Image: WARC AI image

SPT