Brainsuck, or Mining Your Mind
Yet Another Article on Chatbots
After playing with ChatGPT some more in this last week, I agree that it is awesome and a game-changer. It’s been moderately helpful given the right problem space. Even though many of its responses are half-assed, within the scope of certain uses it’s amazing. Wonderful!
OK ChatGPT is Awesome. But.
I’m sure the discovery of oil was accompanied by similar sentiments. But how is this changed game going to be played? Does it just enable more breathless consumption and destruction, or something else? If the past is any guide the most likely outcome is that dopamine-driven exuberance will power creation of a system that produces some goodness, generates mountains of mediocre blather at great expense, and eventually does a whole lot of unforeseen damage.
Your Brain on ChatGPT
An analogy with the petroleum revolution is apt. Your brain is a small pool of something like crude oil. All the billions of other brains on the planet contain the same. Tubes have been inserted into all those skulls, through which giant refineries suck this resource.
This stuff is being refined into fuel to power a robot infrastructure built for creating email memos, logos, advertisements, annual reports, laws, marketing campaigns, clickbait, research papers, and mediocre blogs at ever faster rates.
The output is also fed back into the robot infrastructure where the bots converse and learn from each other, generate unsupervised content, and pump the result to humans to stimulate more production of raw material. This is much like fracking or using natural gas to force more crude from bedrock.
What’s missing from this picture?
There is no feedback from humans to control operational parameters of the bots.
Control over chatbot behavior is entirely in the hands of a small cadre of owners, constrained only by handwringers from the public sector. Feedback goes into a black hole of “Thank you for your input. This helps improve everyone’s experience.”
I tried to turn off the use of personal pronouns in ChatGPT, as anthropomorphizing a bot is dangerous. The response was “I will not use personal pronouns.”
There is no room for human well-being and contact with the Real World.
As used by the hypercapitalist produce-consume complex, the speeding hamster wheel of Human->Machine->Machine->Machine->Human->Repeat leaves no room for creative stimulation, exposure to novel ideas, social interactions, or contact with nature.
This is just a systemic outcome of the main motivation to push widespread adoption of the technology: making it easier for humans to do more X which allows them to do Y faster and feed the refineries more rapidly.
ChatGPT and other models work by constructing mathematical models of the probabilities of word patterns. Anyone who has an intuition for statistics should see that this is a guarantee of mediocrity. “50% of everyone is at or below average.”
I entered a query for ChatGPT to construct the AI-petroleum analogy. It did a reasonable but bloodless job of putting some obvious bullet points together. Then I tried to get output that was more creative and that probed less probable language patterns. You can see the resulting lack of discernment, creativity, spark, unique voice, and humanity in the result at https://pastebin.com/JGMWdcfS
Should you choose to accept them….
Try to get the ChatGPT generator to stop outputting personal pronouns to refer to the app and its output. “I’m sorry, Dave. I have no intention of doing that.”
Check out what’s happening at the world-wide conference on AI safety at Bletchley Park. You could always ask ChatGPT for a summary of the presentations, but I will give you a summary now: “We must do something! Whatever shall we do?”
If someone forwarded this and you found value in it, your subscription fuels my need for reinforcement and helps me create more insightful critiques of the ways of the world. Always Free!