On Strong AI & Robotics

When AI Phones It In

Who Cares About Quality if Costs Are Down?

The latest round of Generative AI with imagery, and also ChatGPT, have renewed the interest in fairness and quality of AI horning in on traditionally human jobs and/or outright replacing humans.

It’s a subset of a basic fear of technology, and has trended every once in a while in the past—hundreds, thousands of years? And the “AI stealing your job” or “robots will steal your job” has come up every so often in the past several decades, often completely alarmist, But I have a feeling these discussions will trend more frequently in the next couple decades especially with AI.

Language translators have apparently gotten a raw deal with AI:

I spent decades of my life learning foreign languages, only to see the translation industry destroyed by AI. The inferiority of the machine translations a few years back did not stop the destruction of the industry. The machine translation cost nothing, and so the price for all translation came crashing down, because the bottom feeders used machine translation. I found myself paid half price to ‘just edit’ (as if it was less work) a translation done by machine which was basically unintelligible so that I had to go back to the original and translate it myself. Most clients, the bottom of the pyramid that kept the industry going, did not care about the quality of the translation. If we expect that clients prizing human made products will save industries we are being very delusional. … the vast majority of clients will go for the process that costs less.

Caroline Kloppert, comment on YouTube (thanks Lauren Wilford)

This is intriguing to me, since it’s so easy to assume that all automation and AI tools that would effectively push humans out of jobs would not change the quality, certainly not lower it, of our products and services.

This is a great analogy for AI art because quick & dirty translation software is an INCREDIBLE asset, and I wouldn’t want to do without it – but it exists in an ecosystem where the cheapest option is incentivized, so experts are put out of work and we all get a shittier product

Jane Waddell

I should not have been surprised. And I’d love to hear of other examples like this.

Transformational Moment?

The next couple of years is going to be one of those transformational moments in technology. Think iPhone, think Mosaic browser.

Reid Hoffman, partner, Greylock at Fortune’s Brainstorm A.I. conference

Perhaps this is largely not an iPhone or Mosaic moment in history, but more one of VHS VCRs in the 1980s or the MP3 format in the late 1990s—convenience over fidelity. And lower cost than the former best and/or only choices as well.

And now something funny: Will A.I. like ChatGPT or GitHub Copilot replace human programmers?

Well, programmers probably aren’t going away too soon even with Copilot. And remember, programmers are a critical part of AI research and deploying AI, so if we put ourselves out of business what are we doing?

And in many contexts, the lowest bidder is going to have the worst outcome. In projects involving software engineering, this could result in having to redo almost everything—making it cost twice as much and take twice as long.

They’re Coming to Get You, Barbara

But for other white collar jobs…

Predicting the future is hard but it seems plausible to me that GPT and software like it could destroy the college wage premium over the next 20 or 30 years.

A lot of white collar jobs basically involve compiling a bunch of information and synthesizing it into a spreadsheet or memo. We will still need humans to oversee this process but a lot more of it might be doable by machines in 20 years.

Whereas a neural network is not going to renovate your kitchen or help you with physical therapy.

Timothy B. Lee

You might be thinking: Fool! It will be even worse because robots powered by AI will do the physical stuff!

But I think you need to consider that timeline of 20 years. Robotics has, sadly, developed slowly especially with real-world unconstrained environments.

An example: You could spend a year setting up a robot to deal with some level of unpredictability in a particular kitchen, doing some specific tasks. Try scaling that to any kitchen anywhere, and do it before funding runs out.