If you are curious about the title of this post “agentic capabilities” it is, according to DeepSeek a Chinese developed AI bot:
“In simple terms, agentic capabilities are what turn an AI from a chatbot (that just answers questions) into an autonomous agent (that actually gets things done for you).”
Sounds good, my own agent, right? The jury is out on that. I started looking at this after reading a bit about Z.ai which is another Chinese developed bot with superior “agentic capabilities” Anthropic’s Claude bot is supposed to be pretty good and Google’s Gemini leads the field in the autonomous agent race (to a beautiful world, the road to nowhere or some form of hell - who knows?
The general public has primarily been using AI bots as chatbots. They can be quite useful for discussing and learning about a variety of topics - novels, history, philosophy, music, art, cars, motorcycles, cooking, movies, TV shows, science, physics, math, etc. etc. etc.
It helps if you have some background knowledge in the area you would like to learn about since that helps you frame questions as well as spot errors and omissions. My favorites are Google's Gemini, the Chinese DeepSeek and lately Anthropic's Claude.
There are many stories of people using chatbots in harmful ways as well. Not to be too pedantic about it but the internet, computers, AI and technology in general are a sharp knife - used incorrectly they can do great harm (he said as the Southwest runs out of water due to human-caused climate change induced since the industrial revolution).
I'm a hobbyist/casual user and don't know what businesses, students, teachers, artists, scammers or YouTube slop-makers are using current AI technology for. I have some suspicions however that what appears to be a general enshitification of various things is due to AI bot use. Specifically scheduling, shipping, internet slop (of course), customer service and a variety of customer interactions.
Fun fact - the rate of "rage-clicks" on the unhelpful chatbots you see on many websites is skyrocketing. You’d have thought we would have learned from the rather unfortunate clippy experience back in the last century - but no.
A couple of examples:
Amazon shipping and scheduling seems to have taken a fairly sharp drop in accuracy lately for things I order. Shipping delays, inaccurate shipment arrival dates once rare are now fairly commonplace. There are multiple potential reasons for this, but I think attempting to replace humans, with AI, plays a role. (*update below)
My medical insurance company nuisance calling me multiple times to “help” me even though I explain I don’t need or want help. In this case it's not AI making the calls but nurses or some sort of medical professionals who are being directed (misdirected) by AI.It was all well meaning but a total waste of time and hard to stop.
I can think of many other examples - AT&T's always awful but somehow actually more awful "customer service" with the aid of AI, but that's boring.
If things are getting worse, I'm afraid the onset of AI "agentic capabilities" is going to lead to an acceleration of falling down.. "Agentic capabilities" is just a fancy way of saying an AI bot can "do" things rather than just chat with you about the characteristics of cockatoos (for example).
At some point in recent history getting software/computers to do special things, custom things, took a certain amount of knowledge and skill.
DeepSeek tells me,
“Before, to hack a server, you needed to learn Bash, Python, networking, and exploit syntax. Now, an unskilled person simply types: "Write a Python script to brute-force SSH logins using this password list, and if successful, install a backdoor." The AI writes bulletproof, production-level code instantly. The barrier to entry for cybercrime has dropped to zero.”
There are kiddie-scripts for some things in this area but without the skill and knowledge you're pretty much flying blind and you are probably about as likely to catch a virus/bug or be the victim of a scam, on the sites you find this kind of thing on - as you are to actually succeed at your nefarious task.
This is all changing with the advent of some of the newer AI bots. You can give them a task, and they will try to fulfil it. They will "do" things. Want to schedule an appointment with your doctor? How about shop around for the lowest price doctor? Just make sure your instructions are correct -you don’t want a Dr. of philosophy or a veterinarian giving you a cheap colonoscopy. How about calling everyone in town with a voice mimicking a well known authority telling them the water supply is contaminated? AI with agentic capability is ready or if not exactly ready very close once you understand how to modify the open-source software.
With “agentic capable” AI you will have the capability for accidentally or deliberately causing hard to quantify damage. After you jailbreak your bot the potential becomes unimaginable. Read about the Anthropic “Project Vend” experiment to get some sense of how well a bot can “do” things.
At this point you may be asking, "we are giving up precious water, adding to greenhouse gas emissions and raising the price of electricity for this?"
There's a million ways this could go. Right now it's an arms race between the white hats and black hats, complicated by the challenge of figuring out who the good guys are. Commercial AI bots have guidelines built in so if you try to get them to do something bad (that the programmers foresaw) they will refuse. With open-source AI bots most of this goes away as hobbyists, criminals etc. will modify the tool to do what they want it to.
If I was betting on Kalshi I'd select one of two competing outcomes - humans will either reject AI's current path - before things completely fall apart or after.
Humans will eventually reject specific usage of AI, including attempting to replace human things like human contact, coordination, working-together, creativity and knowledge creation. Does that happen before the dystopia occurs or after?
My money is on after. Never underestimate the greed, lack of conscience and lust for power of some individuals or the stupidity of a group. Tend your own garden gentle reader, fixing or understanding a human let alone “humanity" is futile. Not infrequently I have no explanation for why I did something - so the likelihood I'd understand why you or a group of youse did (or will do) something seems rather improbable.
There are things I imagine bots can do better than humans - writing computer code and pattern recognition tasks (like maybe a radiologist or detective of some sort would do?) come to mind, But I'm ignorant and skeptical.
It’s reassuring to me to see young people rejecting some of what Jacques Ellul called technique and I hope the trend continues.
I also ran across this YouTube video from Taylor Lorenz that highlights another awful trend in the online world. This is from the video description -
“Pump Fun GO is a new service that lets anyone pay anyone to do anything, and the results are terrifying. From paying people in poor countries $13 to get forehead tattoos to offering $95 for degrading acts, the platform has become a marketplace for human exploitation disguised as "meme coin marketing."
________________________
