AI in 2023 is like the Web in 1993
We don't know how it will turn out
In 1993, the National Center for Supercomputing Applications released Mosaic, the first graphical Web browser.
The Web protocol had debuted on the Internet two years earlier. All the Web had at that point was text and hyperlinks. Outside of universities, no one cared about it. The only businesses that had web sites were a few esoteric tech companies. Microsoft, the software behemoth, did not have an official corporate web site in 1993.
As of 1993, you could not use the web for any of the things we take for granted today. You could not check email on the Web. You could not stream video. You could not shop on line. You could not download a podcast. You could not look up a restaurant in order to browse its menu or look up directions to get you there.
People at home did not have the software (called a Winsock) to connect Windows PCs to the Web. They did not have the bandwidth to download multimedia (a simple graphic might take over a minute to show up on your computer).
Visionaries and entrepreneurs had many ideas for applications to develop using the Web. Most of these dreams never materialized. Others took much longer to arrive than companies hoped and planned for. Many others appeared unexpectedly.
There were many fads and false starts. I can recall many of my own errant predictions. I was confident in 1993 that the real estate industry would be disrupted, and yet 30 years later the process of buying a home is for the most part just as corrupt and inefficient as it was back then.
I did not think that portable, phone-based Web applications would go anywhere. I was clearly wrong about that.
AI is just getting started
I think that artificial intelligence is in that same chaotic, high-potential state in which the Web found itself in 1993. Yes, the field of artificial intelligence in computers has been around a long time. By the same token, as of 1993, the Internet had been around for a quarter century. But the release of NCSA Mosaic did for the Internet what the release of ChatGPT did for AI.
The newest buzzword in AI is “multimodal,” meaning combining text, images, and possibly other sensory inputs as well as data. Today’s text-only chatbots may soon look as primitive as the text-only Web of 1992.
Tim B. Lee’s recent essay foresees powerful results from combining AI’s with cameras.
Until recently, computers needed human help to understand what was in an image. Now computers can glean a ton of actionable information directly from images. And that data can then become an input to other software.
My advice is to set aside time to play with AI’s now. But imagine directions that the technology might take in the future. We’re in early days.
substacks referenced above: