AI, Take The Wheel!

“This [AI] thing–is it a virus, a drug, or a religion?”

Juanita shrugs. “What’s the difference?”
― Neal Stephenson, Snow Crash

If you haven’t heard about the Generative AI revolution it’s likely you’ve been living as a hermit in a cave. It’s everywhere. It feels as if the world has embraced it with open arms, but the truth is that it’s bumpy in the trenches. This is my experience thus far as an engineer and a manager.

First, a disclaimer: I’m not against generative AI technologies. I think they are neither good nor bad. They are merely yet another tool in our arsenal. However, there are some very serious problems that I see associated with the current approach to GenAI.

1. Licensing and Intellectual Property Violations

The use of “public” data for training models has led to gross violations of intellectual property. Companies developing LLMs have made significant profits by essentially stealing content. Just because information is accessible doesn’t mean it’s free to use commercially. There are ways to indicate to web crawlers that data shouldn’t be used for model training, but the respect for these measures is questionable. Then there’s the matter of End User License Agreements. Adobe was caught forcing its customers to grant them wildly permissive license to effectively do as they pleased. Although Adobe promised that they would not use customer content for training, can they really be trusted given their shady practices? Established precedent show companies to really only respect one thing: the need to increase share price and appease Wall Street. However we are starting to see an increase in lawsuits, which I hope establishes legal precedents to prevent such behaviors.

2. Replacement of Human Jobs

Companies often deny that they intend to replace humans with AI, citing economic challenges as the reason for cutting costs. However, a glance at quarterly reports of major tech companies shows significant profits. Phrases like “we need to operate leaner” are often code for increasing share price at the expense of jobs. In my own experience, we’ve been told that if we’re not using AI in our daily tasks, we’re not doing our jobs. This has understandably been met with resistance. Tech workers can see through the corporate rhetoric. This will likely be most challenging in the US, where worker protections are largely nonexistent. The only thing protecting companies at the time of this writing is the incredibly difficult labor market. And by “difficult labor market” I mean “companies have fired a lot of people and it’s tough to get a job.” Companies beware, though, once the market conditions improve there’s going to be a mass exodus from wherever Finance squeezed budgets and shoved AI down everyone’s throats.

3. Rushing Towards General Intelligence

The potential existential crisis posed by rapid AI advancement sounds like sci-fi, but it’s real. The exponential growth in AI capabilities is hard to comprehend. Comparing the improvements in the last year to the last six months shows remarkable advancements. Although we won’t see Moore’s Law-like doubling of capability every year, we will see a compound interest-like effect. The problem is we can’t predict the rate at which AI will grow in ability. This unpredictability is what makes the situation worrisome. We need to slow down and establish protocols and controls to prevent catastrophic scenarios. Unfortunately, between some people’s naivete and Wall Street’s insatiable appetite for share price growth, a cautious approach seems unlikely.

However, it’s not all doom and gloom. There are tangible benefits to GenAI. For example, it helps me significantly in my work tasks. Preparing presentations or reports, which once took days of Excel analysis, now takes hours. GenAI assists in crafting and reviewing emails, checking for tone, and searching through documents and emails. The AI summary and analysis feature in Teams transcripts is excellent. From a coding perspective, the AI’s suggestions are almost eerily accurate, hitting about 80% correctness. This allows me to focus on the more interesting aspects of coding rather than repetitive tasks. It also helps answer questions when I get stuck (these days that’s a lot as coding is not my primary focus). I do believe it can help other engineers in their day-to-day if used correctly. I find that it’s most powerful acting as a cognitive prosthesis.

In conclusion, GenAI is a mixed bag. It offers benefits but has arrived through questionable means, and companies continue to exploit it to cut labor costs while denying it. Don’t believe everything you hear, folks. The truth is out there.

1