May
13
AI: finding the metaphor
Filed Under Uncategorized | Leave a Comment
Imagine the conversation inside the Silicon Valley marketing department. “The headlines are killing us, Herb! While the scientists down the corridor are dreaming of AGI, we have to sell the damn stuff to someone. And with the best will in the world that gets hard when there is a new story every day about AI killing jobs. And it’s one thing when you automate a car plant with robots, quite another thing when you automate a law office, know what I mean Herb? And then when you automate poets and research scientists and photographers and screen actors– well, it gets to a point when it’s just not enough to say that AI will cure the common cold and read x-rays better than any eyes on earth, you just gotta do something. And what do we do, Herb, when users mistrust us, and feel alienated by our products? Yes, Herb, you got it! We change the product names!“
I am not suggesting for a moment, of course, that a conspiracy of intent has taken place in Silicon Valley. Marketing men no longer meet in the dark and private rooms of restaurants off Market Street for very good anti-trust reasons. They use end to end encrypted messaging systems, instead, Which is why my suggested conversation is entirely fictional. But you do not need to be much of a conspiracy theorist to look at the huge costs of creating generative AI, together with the huge valuations created in recent fundraising to put two and two together and get “dotcom boom to dotcom bust“ syndrome. Since three of the largest five big tech players in the US seem to be betting the shop on AI, with generative AI at its core, and since US economic buoyancy, led by the tech sector, has been a critical factor in the recovery of the global economy from the pandemic, we all need to listen carefully. So how are we going to sell the AI that we have today, call it generative or not, call it AI or machine intelligence, or not? The answer my friends will not be written in the wind this time: it will be written in the titles that we choose for our products and services.
Microsoft started it. Copilot was brilliant. Notice how they dropped that horrid hyphen? Were they under pressure from coders, or trademark lawyers, or from advertising agencies? Whatever the case their decision is now binding in the world of CO nomenclature, which is where AI is now covering its tracks. if you have a Law practice AI environment, call it, CoCounsel (but watch out,Thomson Reuters got there first). I hear talk of a copilot in data fusion, a corecruiter in staff selection, a CoDirector in boardroom compliance. There is a reason for this, best told by my marketing colleagues on the West Coast.
“You see Herb, it’s like this. No one likes to feel threatened. Buy our products and they could eat your job! Never a good selling proposition. While we have been putting more and more AI into process for 25 years, so gradually that no one has really noticed, the great generative AI push of the last 18 months has changed everything. We know that auto pilot can fly aeroplanes just as well as people, but somehow we feel comforted that human judgement, or irrationality, can intercede at any point. So let us position the AI beside you, adding value to what you do, increasingly doing what you do, but we won’t be dwelling upon that in the interim. Let’s concentrate the minds of users insteadon small efficiencies, incremental gains, with the machine intelligence, doing the things that it does best, doing scale things that we could not do at all, and gradually becoming enabled as the full service process controller, gradually getting ready to slide into the pilot seat. And that, Herb, is when the nomenclature really does work out. The machine intelligence becomes the pilot, and the user becomes the copilot, only needed in an emergency. And while all of this is going on, Herb, we will have time to deal with a little emergency of our own. it’s called the “where do we get the data from “crisis. You know all these publishers, data, providers and so-called proprietary data owners have been bleating forever that we have been nicking their data and using it illegally? Turns out, it’s not even the right data, and for the most part, there is not enough of it. To make our models work, it is becoming clear that we are going to have to find data, originate data, record data that is out there, but is not yet being collected. And we are having to prepare data for use in different intelligent context, from the cloud to the training set. You see Herb, turns out it’s just like everything else that we do here. We issue the press release on day one to tell the world what we’ve accomplished and how glorious is the new world that we have built, and then wewake up on day two, with the task of actually accomplishing it before us!“