Receive free Artificial intelligence updates
We’ll send you a myFT Daily Digest email rounding up the latest Artificial intelligence news every morning.
Is artificial intelligence best seen as a business accelerant — something that brings new, “smart” features to existing products and provides a lift to tech companies’ revenues?
Or is it a disruptive new platform, with the potential to change the balance of power in the tech industry?
That question has been hanging in the air ever since ChatGPT burst on to the scene. Generative AI — the type that can compose an essay for homework or produce a picture on command — promises to introduce a valuable dimension to many software products. How much more useful would the application you are using be if it could anticipate your next need and produce a creative suggestion?
By offering a whole new way of interacting with computers, however, the technology could also become a platform in its own right. Many applications already plug into OpenAI’s large language models, drawing on them to add AI features. Historically, tech platforms such as this have tended to be very concentrated — whether that involves the software behind laptop computers (Windows or MacOS), smartphones (iPhone or Android) or browsers (Chrome or Safari). How many of the so-called AI “foundation models” will the market support?
Investors, for now, are very much focused on the idea of AI as an accelerant. This week’s earnings calls at Microsoft, Alphabet and Meta were beset by questions about when generative AI enhancements will feed through into higher revenue (and how much AI-powered services might cost to build and run).
Microsoft’s decision last week to slap a $30-a-month premium on the AI features it plans to add to its productivity apps has been the most confident assertion yet about the coming generative AI wave.
But if large language models become a core technology on which many future tech products and services depend, it suggests control of a cutting-edge LLM will be a prerequisite for entry into tech’s big leagues. Microsoft, through its close tie to OpenAI, and Google are the furthest ahead in developing this technology. Can others match them as they race to build ever-larger AI models — and what will it mean if they can’t?
Apple, Amazon and Meta have tended to look at machine learning as an important ingredient in their products, but still only an ingredient — something to improve a recommendations system or provide analytical tools to advertisers. They are now hurrying to give AI a more central place in their operations.
Amazon Web Services seems to have been caught flat-footed by Microsoft, its closest rival in cloud computing, when it comes to running generative AI services. It’s still early days, but Microsoft predicted that AI would add more than 2 percentage points of growth to its cloud business this quarter. AWS is now moving fast to catch up — this week it announced a spate of developments, including offering its customers access to a wider range of LLMs from other companies.
The 3.9bn people who use a Meta app at least once a month, meanwhile, make up a massive potential audience. Chief executive Mark Zuckerberg this week gave a hint of what to expect: a range of AI-powered assistants, agents and avatars would soon populate the company’s services, he suggested, helping users produce better content and increasing their interaction with chatbots.
To support such ambitions, Meta has produced a family of open-source language models called Llama. Amazon is developing its own models, called Titan, and Bloomberg reported last week that Apple had also joined the race.
Zuckerberg told Meta’s investors this week that he didn’t believe there would be “one single AI that people interact with”, and that creators and others would prefer different tools for different purposes. AWS boss Adam Selipsky made the same point in an interview with the Financial Times: there won’t be “one model to rule them all”, he predicted.
It is too early to judge the outcome. In the early days of internet search it was also widely believed that a number of rival services would compete, with specialised search engines filling different niches in the market. Instead, a single, general-purpose search service ended up dominating. OpenAI’s bet on building ever-larger language models is that the same will be true of AI.
If Wall Street’s optimistic predictions prove right, it won’t be long before the first fruits of the generative AI boom start to show up in tech companies’ earnings reports. The ultimate impact on the tech industry will take far longer to become clear.
Read the full article here
