UK’s Competition and Markets Authority Launches Review into AI Foundation Models

From Inside Privacy – Covington:

On 4 May 2023, the UK Competition and Markets Authority (“CMA”) announced it is launching a review into AI foundation models and their potential implications for the UK competition and consumer protection regime. The CMA’s review is part of the UK’s wider approach to AI regulation which will require existing regulators to take responsibility for promoting and overseeing responsible AI within their sectors . . . . The UK Information Commissioner’s Office (“ICO”) has also recently published guidance for businesses on best practices for data protection-compliant AI.

The CMA’s focus is on foundation models – a type of AI model trained on large amounts of data that can be adapted to a wide range of different tasks and services such as chatbots and image generators – and how their use could evolve in the future. The review will focus on three main themes:

  • Competition and barriers to entry in the development of foundation models;
  • The impact foundation models may have on competition in other markets; and
  • Potential risks to consumers arising from the use of foundation models in products and services.

As part of its evidence gathering efforts, the CMA will issue “short information requests” to key players including “industry labs developing foundation models, developers… leading technology firms” and others 

Link to the rest at Inside Privacy – Covington

“Covington” in the source refers to Covington & Burling, an extremely large world-wide law firm, founded in 1913 in Washington, DC., by the two original partners. Covington grew to 100 attorneys by 1960, more than 200 attorneys by 1980. Today, Covington & Burling has more than 1,300 attorneys plus many, many more paralegals, assistants and inside experts in 13 offices, including places like Dubai, Johannesburg and Frankfurt.

One of the more recently-created practice areas focuses on legal issues related to artificial intelligence and robotics. It’s managed by three senior partners located in Washington, New York and Frankfurt. In a couple of weeks, the Artificial Intelligence and Robots group will host its 2023 Robotics Forum which will include presentations on subjects like Regulation of Data in Machine Learning and AI.

Per the OP, lots of non-technical law makers are trying to understand AI and, PG suspects, have no idea how they will or can regulate it in one way or another. Since a great deal of AI can be reached and used via the internet and, PG suspects, that AI can or will soon be able to live in distributed computing environments linked by high speed data connections, the question of what government or collection of governments has the ability to regulate AI usage will be a real hairball. PG suspects Covington and similarly large international law firms want to be exceedingly involved in those sorts of questions.

4 thoughts on “UK’s Competition and Markets Authority Launches Review into AI Foundation Models”

  1. There was a fluff piece on the news, Saying they’re where concerns about “Generative” AI. with vague threats about what it could enable and saying the Australian Government should get out in front with regulating the subject based on how much success they had with Google.

    My thoughts where about the same, sure they can regulate something grandstanding and political but what real affect will it have?

  2. I’m waiting for the desktop versions where one could feasibly train it as an expert system, limited to a specific area. Just turn it on and leave it alone for a week, month, year…

    Who would regulate that? How?

    • Or why.

      And yes, that is coming.
      I’ve already mentioned there already is a version of Stable Diffusion that runs localy on high end gaming PCs. The video card is ridiculously expensive but it is doable today. And, of course, that price will come down.

      https://www.tomshardware.com/news/stable-diffusion-gpu-benchmarks

      (Also, that is for art; text use should be leaner.)

      As usual with tech, by the time the bureaucrats get a handle on a development it has evolved beyond their grasp. To a large extent the “swiss army knife” app like ChatGPT Premium is a tech demonstrator. The real production apps won’t be about “oh, its AI magic!” but instead they’ll be “Xyz app 2024 edition, now with AI”. And no, the fools won’t have a clue what to do.

      Dunno how many out there might have survived the OOP wars of the 90’s but once upon a time the entire computing world was all atwitter over object oriented programming with entire armies evangelizing programming languages left and right, Smalltall vs Objective C vs Logo vs… Hype wars.

      (Steve Jobs Next cube only ran Objective C. He swore it was the best and tried to bring it to Apple when he returned. No luck.)

      For all the hype and sniping in the computer science circles, none of ” the best” languages actually won. Instead the underlying principles and best practices of all migrated into new versions of exiting tools. If anything could be said to “win” the OOP wars, it was C++. At least for commercial software.

      My take, FWIW, is that LLMs will not take over the software industry, much less the world. Hype aside, it’s just a new plumber’s tool–albeit it a powerful one–and people (and companies) don’t buy software because of the plumbing.

      The politicians and bureaucrats are buying into hype and chasing an illusion.

      Let them. Just sit back and watch them chase their tails.

      • Here’s another example of how “AI” is really going to deploy:

        https://www.windowscentral.com/software-apps/slack-gpt-chatgpt-for-slack-now-too

        Slack is a business online meeting and collaboration tool, good for smaller businesses. Second, and slipping, to Microsoft TEAMS. And MS has already added GPT4 to TEAMS since February so Slack has to answer:

        “As found on the official ChatGPT for Slack page(opens in new tab), Slack GPT “combines knowledge found in Slack with the intelligence of ChatGPT, empowering customers with the information they need to move work forward faster.” A few ways in which ChatGPT for Slack empowers users are by:

        Generating summaries to help users quickly catch up on the latest info
        Quickly finding answers after scanning messages and chats
        Drafting messages in just a few seconds

        The idea is that these features will increase productivity and allow users to be more in the loop without taking away from their busy schedules. Chief Product Officer at Slack, Noah Desai Weiss, explained how great of an opportunity it was for Slack and OpenAI to integrate like this.

        “OpenAI has been a great Slack customer, and we’re even more excited for them to be an amazing Slack partner. The ChatGPT app for Slack deeply integrates the power of OpenAI’s cutting edge large language models into Slack’s conversational interface. There couldn’t be a more natural fit. This will give customers new superpowers by helping them tap the collective knowledge of their organization’s channel archives. We’re excited to partner with OpenAI to bring more generative AI powers directly into Slack to deliver productivity efficiencies for everyone.”

        —-

        Its an arms race and companies have to match the new capabilities of competing apps. Or at least pretend to. 🙂

        What the aparatchiks may think won’t factor into it. Even if they could act fast enough. It’ll be all over before they can even draft a position paper.

Comments are closed.