A few decades ago, one of the most popular, wine companies in the US, Paul Masson, hit Advertising gold with the slogan “We Sell no wine before it’s Time’. They even got iconic film director and actor Orson Welles to their TV commercials. Would have made Mad Men’s Don Draper proud.
All this comes to mind as I look at the current rush to sell AI in all its forms at ever increasing prices in these early days of the AI Tech Wave.
Which is fine, if the underlying AI worked as promised. Most AI services today, from Microsoft’s Copilot for everything from Office to Windows and more, to the $20/month+ ‘premium’ chatbot subscriptions from OpenAI, Google, Microsoft, and many others. They’re all promising that today’s AI will meaningfully augment today’s knowledge worker. From coders, to spreadsheet wonks, to market types, to customer support folks trying to do more with less.
And the results from even the latest and greatest LLM AI models from the big ‘hyper scalers’ like OpenAI, Google and others, are providing results on a wide spectrum of results with varying degrees of reliability, accuracy and safety. And measures of evaluating the same still in dramatic flux.
And all of it is happening when the underlying costs of providing ever more ‘hyper-scaled’ LLM AI keeps going up vs traditional software for search and productivity. Yes, these costs will come down over time, but for now, particularly due to the high cost of the critical ‘reinforced learning Inference loops’ in the chart above that I’ve written about extensively. Their costs for now, are relatively unbounded, with relatively bounded pricing.
This selling of ‘software before it’s time’ is not new in the tech world, and through various tech cycles. Companies like Google and others are famous for deploying ‘Beta’ software and services for years before the underlying tech was ‘baked’ enough to be an official product. Google with Gmail is a famous example, shedding its ‘beta’ status after half a decade.
I’ve long discussed how ‘self-driving cars’ are the ‘canary in the AI coalmine’. Recent examples of course include Tesla, Apple and others investing and striving to achieve ‘Level 5’ Full Self-Driving and yet delivering barely in the upper founds of Level 2.
Even Tesla and Elon Musk, after promising ‘FSD’ of Full-Self Driving for years, is barely at that level. This despite the current vaunted transition of Tesla’s FSD from version 11 to version 12, still in ‘Beta’. One can watch many on-going video reviews of FSD 12.3.1 on the road in real-life conditions. Here’s one short example for a taste.
Tesla’s FSD 12 software promises to transition Tesla from traditionally hard machine coded self-driving to ‘Generative AI’ model driven self-driving based on the immense real-time data from almost two million Teslas on the road. Much of that increasingly running on Elon’s heralded ‘Dojo’ AI super-computer infrastructure.
The reviews are promising, but generally viewed as a work in progress. Yet Tesla sells ‘Full Self-Driving’ FSD packages at $12,000 one-time fee, or a $200/month service fee. Tesla has approximately 400,000 FSD customers on an overall Tesla user base of almost three million. After over a decade of promising ‘full self-driving’ and indeed, robotic taxi capabilities to eager, early buyers.
As AI transitions from its early days from experiences like ChatGPT to much more in ‘multimodal’ capabilities with images, video and more, we’re in for a similarly long transition of AI software being bought by eager, early buyers long before ‘it’s time’.
And we’ll continue to see this as AI services show up in other forms like AI Pins, standalone handhelds and even ‘Smart Glasses’ on our faces from capable companies like Meta , Apple and others.
Before the AI does what the end users, business and consumer, think it will really do. And because we’re in a world with over four billion smartphone users worldwide, the ‘early’ buyers and users of AI products and services will be measures in the billions and more.
Far before AI hits its mainstream stride. A lot of AI software and related services to be sold before its time.
Indeed, tens of billions of dollars’ worth worldwide, before it’s all really beyond ‘Beta’.
As we ‘Wait for it’ in terms of the AI products and services that we all aspire to, we’ll have to settle for drinking much ‘AI wine’ before its time. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)
Loved reading