#213: Rise of the robots

Valve sounds a welcome note of caution amid the generative ML goldrush.

There’s a genre of tweet that’s very common whenever some divisive new technology comes along. Perhaps you’ve seen it: a selection of newspaper clippings dating back to the early years of the internet, in which a variety of esteemed publications proclaim the world wide web to be nought but a passing fad. New-tech evangelists trot this out to try and shut down criticism, to silence the naysayers. You luddites were wrong once, they say. You will be proven wrong again.

Obviously, this does not quite stack up. I am sure we could all reel off dozens of pseudo-disruptive new ideas from years gone by that invited a lot of scepticism, and later went on to justify it. (Insert your own N-Gage/3DO/Brexit joke here.) But in fairness I do think some of the clamour around the rapid rise of generative AI — which in accordance with the Hit Points style guide will be henceforth referred to as ‘machine learning’, or some variation thereof — is being somewhat overstated. I think a lot of people have watched Terminator 2 a few too many times, and see ChatGPT as Skynet’s portentous opening gambit. And we must also acknowledge that the tone of the conversation is largely driven by the media, which is hardly an unbiased observer since its trade will be one of the first against the wall when the machines formally rise up against us.

In fairness to my journo colleagues, from their perspective the war has already begun, and the robots have already claimed plenty of territory. Late last year CNET quietly began publishing AI-generated articles; covering some pretty critical topics, among them personal finance, they were riddled with errors — and, hilariously, compelled the human editorial staff to unionise. (CNET, undeterred, pledged to plough on with the endeavour.) Earlier this month Gamurs Group, owner of a raft of websites including Dot Esports, Destructoid and The Mary Sue, advertised for an AI Editor, a newly created position that would oversee the AI-powered production of up to 250 articles per week. Gamurs made this move just three months after laying off 40% of its workforce, blaming “operational inefficiencies.” The job ad was taken down amid the backlash, but I feel pretty confident in saying the position is still very much a thing.

GenML’s implications for journalism are troubling enough; its rapid impact on the wider internet, and the tech industry more broadly, feels even bleaker. Google and Microsoft are redefining search around the technology, pushing the traditional titans of SEO down the rankings and the ad-dependent publishing industry ever closer to the financial abyss. As The Verge noted recently, machine learning is killing what it calls ‘the old web’, and replacing it with something that is somehow even worse. This is real and it is happening, and it is moving much faster than almost anyone could have imagined, regardless of their respective levels of luddism or advocacy for the technology itself.

However, the key word there is ‘almost’. I think it’s fair to say that those who could have seen it coming might have done a little more to roll the wicket for it. Last month Sam Altman, CEO of OpenAI, the San Francisco startup behind ChatGPT, used his appearance before a US Senate subcommittee to call for urgent regulation of automated #content creation, and the goldrush it has sparked. “I think if this technology goes wrong, it can go quite wrong. And we want to be vocal about that,” he said. “We want to work with the government to prevent that from happening.” Sam! This is lovely stuff, but I think perhaps you have come to this realisation a little late, no? Perhaps we might have had this conversation before your company released a machine-learning model that poses an existential threat to the quality of the global internet, and has creative professionals the world over fearing for their careers and livelihoods? Sam? Are you still there?

I guess not. But if Sam hadn’t just run off giggling with hundred-dollar bills flying out of his pockets, I am sure he would agree there is no putting this thing back in the box. Generative machine learning is here, whether we like it or not, because there is too much money to be made from it by people who sold their scruples long ago.

Speaking of scruples, here’s John Riccitiello! Hit Points is no great fan of Riccitiello, head honcho at Unity and the only game-industry CEO to dress like a Final Fight midboss, but credit where it’s due — he knows a good bandwagon when he sees one. Earlier this week Unity announced two new machine-learning tools: Muse, an AI assistant that’s a sort of high-end Clippy for developers, and Sentis, a far more ominous offering that “enables you to embed an AI model in the Unity Runtime for your game or application, enhancing gameplay and other functionality directly on end-user platforms.” For those of who you are understandably struggling to parse that word salad, Sentis enables you to generate game assets with machine learning, rather than paying a human to do it. As part of this, a host of ML content providers were added to the Unity Asset Store.

Unity’s developer community were naturally unnerved by this, wary as they rightly are of genML’s obvious implications for their future employment, housing and healthcare prospects. Kotaku rounded up some of the responses to the announcement, and the tone of them is pretty much what you’d expect. “Hi, what dataset was this trained on?” asks one developer. “Is this using artwork from artists without their permission? Animations? Materials? How was this AI trained?”

This cuts to the core of the issue. There can be no ‘artificial intelligence’ without human input, and the technology that threatens these developers’ livelihoods only exists because it has been trained on the work of them, their colleagues, and their creative peers. Surely Unity will have recognised this, and done its due diligence? Surely it ensured everything was above board before hitting the submit button? Hahaha, no. Within 48 hours the company had announced the delisting of one of the Asset Store providers for violating its T&Cs, presumably over copyright concerns. “We are actively reviewing all verified solutions providers in the AI category,” it said, “as we remain committed to upholding the quality of our ecosystem.” Ah, yes, reviewing. Good. Perhaps a little previewing might have been in order, no? I suppose they are merely following Altman’s example, reaching pointlessly for the stable door as the horse disappears over the horizon, but still. Come on.

Thank heavens, then, for Valve, which was this week revealed to have been quietly banning games that use ML-generated assets. Hit Points chum Simon Carless, consultant and author of the excellent GameDiscoverCo newsletter, shared an email from Valve to an anonymous developer whose game “contains art assets generated by artificial intelligence that appears to be relying on copyrighted material owned by thirdparties. As the legal ownership of such AI-generated art is unclear, we cannot ship your game while it contains these AI-generated assets.” A follow-up tweet by indie developer Jake Birkett suggests Valve is taking the same line with text-based #content.

Now look. I realise Valve may not be acting solely out of principle here, or even out of principle at all. The company made a similar decision when it banned blockchain games back in 2021, and in both cases the vibe I have is of a company that reckons something sounds like more trouble than it is worth, and is opting to simply stay out of it. In blockchain’s case, it was about the likely need for a customer-service function to deal with players complaining about being scammed, or their hard-won NFTs plummeting in value. With genML, Valve is clearly most concerned about the potential legal ramifications.

It seems Newell and co have weighed up the likelihood of getting caught up in a class-action lawsuit. brought by disgruntled creatives who believe the machine-learning nobbers have been scraping their work for inspiration — something that seems pretty inevitable, if not necessarily imminent — and figured, nah, can’t be arsed with that. Fair enough; for once, I reckon the end justifies the means. I am relieved to see someone taking a stand, and a big, successful company recognising that some things are more important than money. The first of many, you’d hope, though I’d caution against holding your breath.


  • Niantic, developer of Pokémon Go and a host of games that are broadly similar to Pokémon Go but are nowhere near as successful as it, is shuttering its LA office, laying off over 200 people, closing its NBA game and cancelling its Marvel one. Looks very much like the wheels are coming off over there, doesn’t it.
  • I had no idea that indie publisher Tinybuild was publicly traded, I must admit, but its stock took an absolute hammering yesterday after it released a rather lukewarm trading update. It missed its revenue forecast by a mile, lowered its projections for the next couple of years, and announced the departure of its CFO; shares fell by as much as 80% as a result. Cripes.
  • Rumours have been swirling about this for a while now, but Xbox Series S is definitely giving developers a headache. Larian Studios reveals it has struggled to get Baldur’s Gate 3’s split-screen co-op working on the entry-level console. Since Microsoft insists on feature parity between it and Series X, the game currently has no release date on Xbox. Less than three years into the Series generation, one images this sort of thing will only grow more prevalent over time.
  • Sony blocked a planned PlayStation release for Roblox over fears it was harmful to children, according to a filing in the FTC vs Microsoft case. However, Jim Ryan says that Sony has since “reviewed those policies and relaxed a little on this… We have been conservative for too long.”
  • Canada’s competition regulator has said that it reckons Acquisition Blizzard is anticompetitive. Microsoft quickly hit back by pointing out that the regulator missed a deadline and is therefore no longer allowed to complain, so ner. There’s been plenty more Activision fun over the last couple of days but I’ll save that for next week, if you don’t mind, when the judge in the FTC case is expected to pass down her ruling.

That’ll do! Gosh, I hope today’s edition was alright — my head really isn’t in the game today, and this sure is a knotty subject. Have a great weekend, whatever it has in store, and I’ll see you all next week.