By Jack Cumming
You’ve read that artificial intelligence (AI) is consuming massive amounts of power and that the hunger for power may bring AI’s potential to its knees before we can reap its benefits. Environmentalists, perhaps also spooked by AI’s potential workforce impact, point to the power drain as a reason to rein AI in. Let me take you back in time to set that into perspective.
The Breakout Computer
One of my first jobs after college was with New York Life, which was the proud owner of an IBM 650 computer. I remember, as if it were yesterday, being allowed into the computer room to get a glimpse of this massively computational genius. Of course, we didn’t then call computing AI, but it was AI nonetheless. The idea of AI wasn’t new. Actuary Edmund Berkeley published a book in 1949 titled “Giant Brains Or Machines That Think.”
The IBM 650 could do tricks that had eluded us, no matter how well we had mastered our Friden and Monroe calculators. Those desktop machines were themselves still considered magical, even though they were fully mechanical and produced fascinating rhythms when keyed just right. But the IBM 650 put those calculators to shame.
Remember Raised Floors?
What struck one on walking into the computer room was the chill in the area and the raised floor. We could glimpse that there were cables strewn everywhere under the floor. The chill in the air was to keep the computers cool. After all, the IBM 650 was filled with vacuum tubes. Miniature vacuum tubes were then the latest thing. They replaced much larger, older vacuum tubes.
Most people today have never seen a vacuum tube. They seem incredibly old-fashioned, but those miniature tubes were a big deal in their day. Every vacuum tube had a filament that glowed like fire. They each gave off a great deal of heat. Remember when incandescent light bulbs gave off heat?
Imagine thousands of vacuum tubes clustered in a computer. The power demands to heat those filaments and then to cool the room was enormous. The IBM 650 held roughly 2,000 vacuum tubes and 3,600 crystal diodes.
Parallels With AI
The IBM 650 was a huge success, with runaway adoption very similar to the adoption of AI today. The market projection was that 50 IBM 650s might be sold, but eventually over 2,000 were installed. Now, think of how much power that installation demanded. The thinking at the time was not dissimilar to what we are reading today about the power demands of AI.
The salient takeaway is that technology never stands still, and we should all be thinking beyond the next corner. Today, we all hold much more computing power than those IBM 650s on our laps as we happily type away while talking on an equally powerful smartphone. We don’t need a dedicated power plant for that. We carry all the power needed with us wherever we decide to work, even in the air while flying around the world.
Embrace the Future
Who would have dreamt that LED lighting would bring down the consumption of power dramatically, just as AI and electric vehicles began their time of high power consumption? How long will it be before AI is as power-efficient as a modern laptop or smartphone? Let’s keep moving forward to reap the benefits that are ours for the taking.
