For the technology sector, 2023 can be characterised as the year of two extremes. There is unprecedented innovation as the exciting potential of artificial intelligence unfolds, but it is also the year of unprecedented regulation and it’s a difficult line to tread.
In the UK alone there is the Online Safety Bill looking at content, the Digital Markets, Competition and Consumers Bill examining competition and an AI summit in Bletchley Park in November looking at safety.
Europe is on a similar path. It has passed the Digital Services Act and the Digital Markets Act when it designated the usual suspects of Alphabet, Amazon, Apple, ByteDance, Meta and Microsoft as so-called “gatekeepers” of the digital economy, their size making them subject to more stringent rules.
In addition, there is a seemingly endless stream of investigations from various bodies into Big Tech’s behaviour. Yesterday, Alphabet made a last-ditch effort to overturn a $2.6 billion fine from the EU over anti-competitive actions related to its shopping service. Ofcom is looking into the market for cloud computing. The Competition and Markets Authority is looking into whether Apple has a monopoly on the App Store. The list goes on.
Morning and midday updates on financial and economic news from our award-winning business team.
The striking difference between the action on AI and all of the others is just how pre-emptive it is.
With AI, unlike social media, watchdogs are trying to get ahead of the game. The attempt is admirable but they have a Herculean task. It was summed up this week, when the CMA published a report about the risks of so-called foundation models (FMs) to business and society. These are a type of AI trained on vast tracts of data, such as those sitting behind chatbots or image creators.
It found that Big Tech, those “gatekeepers” if you will, is likely to continue to dominate the increasingly lucrative market for AI, in what the regulator called a “winner-takes-most” outcome. It is very hard to envisage what the CMA or anyone else can do about that.
To train these models, huge amounts of computing power is needed and a more powerful type of computer chip than is readily available. Then there is the expensive infrastructure — the data centres and servers that Big Tech already has. Those companies without their own data centres have to turn to cloud service providers, who are Google, Amazon and Microsoft.
The better the data, the better the model. YouTube is suggested as one deep pool of information for models, it is of course owned by Google.
Other serious players are also tied into Big Tech. They include OpenAI, backed by Microsoft; Inflection, backed by Nvidia and Microsoft; and Anthropic, backed by Google and Salesforce.
The report explained that “a handful of large technology companies possess substantial compute resources, including AI accelerators, enabling them to pre-train FMs in-house quickly and efficiently”. The advantage is clearly with the incumbents.
Meta Platforms is bucking the trend, winning praise from within the industry for offering its models on an open-source basis, although this is not without controversy.
There is a real danger Big Tech could squeeze out smaller firms, or as one industry figure said, “not let them in in the first place”. And forget businesses for a minute, what about university labs?
One British scale-up AI company, CloudNC, said the answer is to find a niche. “If you aren’t Big Tech, then you can’t develop a solution without an obvious business application — you need to be razor-focused about what your AI will be used for and how it will make a difference for customers, or you’ll burn a lot of money with no reward.”
The CMA said it is possible that there could still be something on offer for smaller businesses and that “innovation, community collaboration, and emerging technologies could potentially disrupt the existing dynamics”.
It concludes that the market can be more competitive, as long as developers can access everything they need on fair commercial terms. But it is difficult to see quite how this will happen without serious intervention.
Katie Prescott is Technology Business Editor of The Times