Stupid is as stupid does

The Internet is both a vast collection of knowledge and a massive sewer soaking up the worst of humanity. So perhaps we shouldn’t be surprised by either the potential of artificial intelligence built upon that knowledge, or the cringeworthy moments such technology can generate.

Still, I found the summary of a two-hour conversation between a New York Times journalist and Bing’s A.I. chat striking:

You should read the entire transcript of the conversation, available here. It is remarkable. It is awkward. It is surprisingly… human? It also seems to fall back on some of the basest desires common to a significant portion of online discourse, be it between humans or bots: It’s surprisingly amorous and prone to violent fantasies.

But my overall takeaway is, the conversation sounds like a chat with a fifth grader. Yes, sure, a fifth grader who is surprisingly amorous and prone to violent fantasies. But even in those moments the real noteworthy thing about the declaration of love, and the step-by-step guide to how the AI could, if it wished, cause havoc, is how immature the speaker appears.

Now of course, saying “it isn’t fully there yet” is pretty naive, shallow criticism to level at a nascent technology. The response is obvious, and valid: “Sure, it isn’t perfect right now, but can’t you see what I can become?”

Perhaps it will. But I think the transcript also points to some of the limitations that will eventually stand in the way of the AI-revolution that some are forecasting.

And now a brief (paid) endorsement:

To the extent that this content is intelligent, it is so because I use tools like Koyfin to make sure I have the best charts, data, analysis, and transcripts available. Koyfin provides a lot of the same tools you get with a Bloomberg terminal or Capital IQ at a fraction of the price.

Interested? Click the button below to get 10% off your first year using Koyfin.

I caught a lot of flak from some readers for saying that AI looks like an incremental step forward, not a revolution. Step away from the sheer wonder that this chat was even able to happen, and that a computer was able to engage with a Times writer for nearly two hours, and there is a lot contained within that would suggest incremental progress from here.

Consider:

1.) AI is only as good as its inputs. This is the classic “garbage in/garbage out” problem. AI will work best if it has as complete of a knowledge base as possible. But adding that vast universe of good data indiscriminately means also adding a lot of the darker, more seedy corners of the Internet as well. You can solve this with human guidance/ supervision. But that greatly adds to the complexity, and reduces the power of AI as the next big thing. You can perhaps solve it with AI! Better filters! But you still come back to the question of the humans programming the filters. We will walk blindly into this future hand in hand with our robot friends, not with them on, um, autopilot.

2.) The last 10% is the hard part. It’s relatively easy to sound like a fifth grader. It is a lot harder to sound like Albert Einstein. Even if we set our expectations well below Albert Einstein, we should not underestimate the complexity of the adult human mind. To extrapolate the relative ease of getting from zero to what we have today and assume future upgrades will be just as straightforward is a grave mistake.

We see this in autonomous vehicles. It is relatively straightforward to program a car with the right sensors to stay in a lane, or even exit off a highway. It is significantly harder to get that car to react to a drunk driver swerving across the road, or to navigate a lane-less country road. It sounds obvious with autonomous that it would be foolish to say, “if we can do the straightforward stuff, the hard stuff will be equally difficult.” This new chat AI is no different. Progress from simple will be orders of magnitude harder than what has been done so far.

We’ve seen it elsewhere as well. Alexa is amazing! Alexa is the stuff we dreamed about twenty years ago. And these days you can find as much complaining or mocking Alexa’s mistakes as you can people amazed at what it can do. That last 10% is hard!

So what does this mean for investors?

It means we will be well-served by avoiding the hype. There is a clear path for incremental progress from here, and real use cases for how AI can help today and will be able to help more in the future. But that help will require a lot of human handholding along the way. AI is more of a productivity enhancement tool than a whole new way to do business. That is, almost by definition, incremental progress.

It is also a reminder of the near-term downside risk. The Bing chat to its credit did a good job of quickly censoring its most devious plans, and keeping the cringe-factor on its confessions of love to a PG rating. Not all will be so lucky. With so many testing these systems, someone at some point is going to get hit with a massive headline risk.

Readers of a certain age will likely recognize the title of this post as a quote from a 1990s Tom Hanks movie. To stay with a theme, investors need to recognize AI as the character “Jenny” from that movie.

It is fantastic. It is mesmerizing. But it is bound to break your heart.

Disclaimer: Fits and Starts DOES NOT provide financial advice. All content is for informational purposes only. Stocks mentioned are as reference only, and a mention should not be interpreted as a buy or sell recommendation. The author is not a registered advisor or a broker/dealer. DO YOUR OWN HOMEWORK. The information contained within is not and should not be construed as investment advice, and does not purport to be.

No statement or expression of opinion, or any other matter herein, directly or indirectly, is an offer or the solicitation of an offer to buy or sell the securities or financial instruments mentioned.