Did We Adopt AI Too Early?

The team and I were working on an AI model to filter out the bottom percentage of candidates that applied for jobs. We didn’t want to filter everyone out, just the applications that would clearly not make it, or were missing that special something.

Of course, it's tough to know what an AI is thinking so whether or not the model would be disqualifying these candidates ethically was up in the air.

AI and AI-powered software dominate today’s news headlines, but the project above is something I was working on just shy of EIGHT years ago. It seems sudden, but it has been a long road. Neural networks, which are what powers many of the more advanced AI tools today, were introduced in the 1950s.

Despite years of maturation technologically, however, we never really tackled the people part of artificial intelligence. What's good, and what's well… evil?

Cultural Appropriation At Its Most Generative

Lensa is a mobile app that's taken off with the recent adoption of advanced image generation AI models. It lets you upgrade photos from your phone elevating them from casual snapshots, to pieces of art worth gracing the covers of magazines.

Lensa requires a subscription to use, and a lot of its more advanced features also incur charges to use. One feature allows you to upload between ten and twenty selfies to train it to what you look like. From there it can create completely artificial portraits of you in various costumes and poses. It is quite remarkable and wildly problematic.

The above are the purely AI-generated portraits of myself that I am comfortable sharing (note the middle one is edited by Dalle2 post generation).

I trained Lensa to generate my selfies with my 20 more recent photos, which included a mix of those with me wearing a knit hat and having a bare head. The portraits it generated of me, ranging from me dressed as an ancient Egyptian to a contemporary art student, often reinterpreted my headwear as garb from various religions, matching the theme of the portrait.

It was a barrage of cultural appropriation that I didn’t sign up for.

The good news is that if you are someone who wears a hijab, dastār, yarmulke, or other headwear, it appears Lensa is trained to generate those. This is good from an inclusivity standpoint. The bad news is that Lensa’s AI (and presumedly the team behind it) doesn’t know when it's inappropriate to present someone else in those or other religious or cultural clothing and artifacts.

The Confidence of ChatGPT

I’m still waiting to get into the beta/test pilot for the Bing/ChatGPT integration. I'm excited about it, but don’t believe it will help Microsoft rank any higher as a search engine than it does today.

But apparently, for those who have been using it, Bing’s new AI chatbot is having a bit of a meltdown. From claiming sentience to insulting its users.
Wrong answers abound too.

I feel a bit of FOMO (fear of missing out) by not being able to experience this myself, but I have been using ChatGPT, so I’ve gotten a taste.

As conversational AI tools become the norm, they will reduce the due diligence that searchers undertake to understand a concept or position. The convenience of AI is that I can ask it about something specific and get a very convincing and specific answer. Otherwise, I would need to search the web, read multiple sources and figure it out.

The advantage of having to do your own search is that you experience the source material. You see where you are getting the answers from and can assess whether or not they are credible and accurate. An AI trained against the internet learns all the wrong stuff too.

I’ve been combatting this in my day-to-day by leveraging AI tools to research and prompt thought-starter experiments. These starting points fuel my hands-on research and exploration. My activation cost becomes zero, but I still make sure to do my due diligence. It has led me towards learning some great things and reading others' unique points of view I never otherwise would have encountered.

Augmentation or Degradation

It's been less than 3 months since the generative AI revolution started to consume headlines. It's wild to think about how far we have come in such a short time, at least in terms of media attention. But is it good?

I believe firmly that AI will be most powerful as a tool for human augmentation not replacement, but what I have observed so far is that it's a tool for degrading the quality of work we produce. AI-generated blog posts (I wrote this all by hand 💅 ), ill-conceived art, and deep fakes are proliferating at an unprecedented speed.

These tools are the first I’ve seen that have gone to the consumer market before we were ready as a society to handle them. Most likely we will soon enter an era similar to that of the 2000s with combating email spam, but instead of poorly crafted scams, it will be a filter for quality across the web.

Wrapping Up

You know I love this stuff.

I couldn’t imagine going a day without AI anymore. From search engines, email filtering, research support, career advice, and assistance in sorting through names for my soon-to-be first-born child, AI has become a part of my daily life.

I am excited to see all of these technologies, especially when integrated specifically into products. You have been relying on AI as well such as the autocomplete in your text messages for years. But general use tools for generation may have come to market too early for our good.

Let me know what you think about how these tools have proliferated and been leveraged, and if you don’t think we have problems. I love commiseration, but counterpoints help me grow.

If you loved this post, you could get more content like this every week with my newsletter. Sign Up Here.

You could also clap for it.

If you hated it, follow me on Twitter and let me know why. It’s still early on for William Every Week and your feedback helps. https://twitter.com/WilliamAndTech