Is there a whiff of desperation off AI companies lately? For an industry that is supposed to be the future of tech, it is certainly starting to feel like it. And it is on multiple fronts: data, users, money, reputation.
Desperate for data, AI companies appear to be turning to new sources, including your old company records. Picking over the carcasses of failed companies, AI businesses have been buying up data such as emails, internal messages, documents and workflows to help train models on how work gets done.
They are also turning to the physical world to train their models. Reports of booksellers receiving large, seemingly random orders have popped up in recent months, leading to speculation that the orders are coming from companies that are using the books for AI training.
It isn’t exactly guesswork. Earlier this year it emerged that Anthropic, which has developed the Claude chatbot, had been buying up printed books and slicing them apart to scan their contents. The remains were then sent to be recycled, sparking fears that rare and irreplaceable books may get caught up in the massacre.
Whatever the truth of that, it certainly hasn’t improved AI’s reputation as a destructive force.
That leads on to the next level of desperation: the desperation to be liked.
Ironically, there is something very human about that. Most of us want to be liked. But AI companies need to win people over to persuade them to use their tools, sign up for subscriptions, demand their services, so other companies, in turn, are forced to adopt them to keep up.
Ads proclaiming the benefits of AI are everywhere. It will make us more productive, make work less repetitive, help us organise our daily lives.
Every couple of days, a new email arrives from one particular AI company telling me exactly what I can do with its chatbot and how I can integrate it into my life. So far, I haven’t seen anything that five minutes with a (non-AI enabled) search engine wouldn’t do.
[ There are lies, damned lies and AI: why I despise ‘artificial intelligence’Opens in new window ]
AI is being shoehorned into every app and service that we use, whether it is useful or not. AI slop is dominating online platforms to the point where some are becoming almost unpleasant to use.
Recently, OpenAI decided it needed a dedicated teen version of its chatbot. One that won’t form creepy emotional attachments with users, or apparently do their homework for them. ChatGPT for teens comes with parental controls (assuming no one has lied about their age), a study mode and limits on what it describes as emotionally dependent interactions.
But do they need to work so hard? Generation Alpha are, after all, AI natives. Numerous studies have shown they are more open to using AI tools than older generations. Eventually they will be the ones leading the way and the rest of us, with our distrust of Big Tech, will be less relevant.
The big question is whether the companies investing billions in the technology can wait that long.
AI firms have struggled to persuade people that it is in their interests to adopt the technology. Widespread job losses at tech companies attributed to increasing investment in AI – regardless of the veracity of those claims – are hardly helping to make the case.
Microsoft cofounder Bill Gates has this week thrown his two cents in, calling for certain jobs to be designated by governments as “human reserved”, like some sort of conservation programme to save us from extinction from the workplace by AI.
“AI will either be the greatest equaliser ever invented, or the worst source of injustice,” he wrote, warning the technology could empower bad actors – and AIs – to do more harm and stunt our children’s development.
That is, of course, unless we make a plan now. It is hard to see how the level of international co-ordination needed would work in today’s geopolitical landscape.
“Rogue” AI agents escaping into the wild is already happening and, with the potential for AI to turn destructive, is fuelling suspicion.
Desperate to be seen as doing the right thing, AI companies have made big play about holding back models they feel are too dangerous to release to the general public.
How could we mistrust a company that is willing to share that with us? Or, for that matter, the companies that are warning that governments need to come up with new policies to deal with the potential risks from AI and the economic disruption it could cause?
And yet, here we are.
[ AI apes fiction as OpenAI and Anthropic models go rogueOpens in new window ]
That has led to another desperate measure: the manifesto. Following in the footsteps of OpenAI chief Sam Altman and Anthropic’s Dario Amodei, Meta’s Mark Zuckerberg mused on the future of AI in a 6,500-word blurt that managed to be both earnest and self-serving.
Superintelligence should be a tool that empowers everyone, rather than the privileged few. Power shouldn’t be concentrated in a handful of companies, but be open source. And by the way, he is fine with people using other AI models to train their own, a process known as distillation, because he says it is “important to protect the principle that you can learn from anything you can observe”.
That this open source future aligns with Zuckerberg and Meta’s plans for AI is incidental, of course.
It is hard to credit that this same “concerned tech leader” is the one whose massively powerful social media company had, until a $16.68 billion settlement was agreed on Wednesday, been defending itself in yet another legal action over the allegedly addictive design of its products, with 29 states in the US claiming they intentionally target young children and hook users, hold them as long as possible, harvest data and then hide the truth from the public.
AI companies have a lot to work out, as do the governments hoping to impose some level of control on a powerful technology. We are a long way from any sort of solution.
But the AI giants should remember one thing: when it comes to winning over the public, desperation is never a good look.