How to "appreciate" AI

Three years ago, crafty marketers apparently trying to sell a film set aside July 16 as "AI Appreciation Day" — and now every year this date brings a wave of lame press releases and half-hearted events.

Why it matters: The celebration's origins may be dodgy, but it's a good occasion to take stock of where the generative AI revolution stands, with all its strengths and weaknesses.

The big picture: GenAI does have promise — it's just not a panacea.

• It has shown it can dramatically help generate code, marketing materials and annual performance reviews.
• But with many of the use cases it's been pressed into during the current AI mania, humans still need to be in the loop for fact checking, quality control and common-sense oversight.

To truly appreciate AI today, we have to recognize its possibilities but never lose sight of its limits.

• The technology is prone to making up answers to questions — the so-called hallucination problem.
• GenAI, at least as it's implemented today, comes with a massive carbon footprint that threatens to derail the climate progress many tech giants have touted.
• The technology brings a high risk for further income concentration. Computers taking more tasks from human workers threatens to lower wages and kill jobs.
• GenAI produces inadvertent misinformation while also significantly lowering the cost of generating intentionally false information.

Even with efforts to make AI more fair, generative AI systems encode all of our human biases.

• Existing data overrepresents the experiences and values of men, English-speakers, heterosexuals and those whose gender conforms to the one they were assigned at birth, to name just a few categories.

Between the lines: Appreciating AI also means recognizing the role humans play in making genAI possible.

• That includes the low-paid contractors around the world who provide the feedback that makes the technology closer to usably reliable.

• It also includes the humans whose art, music and words are at the heart of training the systems that spit out music and drawings and poetry on demand.
• Today's genAI overwhelmingly fails to credit, let alone compensate, those whose contributions serve as its foundation.

Our thought bubble: Axios tech editor Megan Morrone put it well in our Slack discussion of this AI "holiday."

• "I refuse to appreciate AI until AI can appreciate me," she said.

What's next: The core technology continues to improve, both with larger datasets and improved reasoning capabilities.

• What's more, humans are getting better at integrating genAI into their own work — using it to handle one step in a larger business or creative process, rather than just trying to hand off an entire job to ChatGPT.

The bottom line: Making the best use of generative AI requires an appreciation of the technology. But appreciation does not obligate a blind embrace.