A Revised AI Glossary for 2026

January 5, 2026

Another dinobaby post. No AI unless it is an image. This dinobaby is not Grandma Moses, just Grandpa Arnold.

I have a lot to do. I spotted this article: “ChatGPT Glossary: 61 AI Terms Everyone Should Know.” I read the list and the definitions. I have decided to invest a bit of time to de-obfuscate this selection of verbal gymnastics. My hunch is that few will be amused. I, however, find this type of exercise very entertaining. On the other hand, my reframing of this “everyone should know” stuff reflects on my role as an addled dinobaby limping around rural Kentucky.

Herewith my recasting of the “everyone should know” list. Remember. Everyone means everyone. That’s a categorical affirmative, and these assertions trigger me.

Artificial general intelligence. Sci-fi craziness from “we can rule the world” wizards

Agentive. You, human, don’t do this stuff anymore.

AI ethics. Anything goes, bros.

AI psychosis. It’s software and it’s driving you nuts.

AI safety. Sure, only if something does not make money or abrogate our control

Algorithms. Numerical recipes explained in classes normal people do not take

Alignment. Weaponizing models and information outputs

Anthropomorphism. Sure, fall in love with outputs. We don’t really care. Just click on the sponsored content.

Artificial intelligence. Code words for baloney and raising money

Autonomous agents. Stay home and make stuff to sell on Etsy

Bias. Our way is the only way

Chatbot. Talk to our model, pal

Claude. An example of tech bro-loney

Cognitive computing. A librarian consultant’s contribution to gibberish

Data augmentation. Indexing

Dataset. Anything an AI outfit can grab and process

Deep learning. Pretending to be smart

Diffusion. Moral dissipation and hot gas

Emergent behavior. Shameless rip off of the Santa Fe Institute and Walter Kaufman

End-to-end learning. Update models instead of retraining them

Ethical considerations. Pontifical statements or “"Those are my ethical principles, and if you don’t like them… well, I have others."

Foom. GenZ’s spelling of the Road Runner’s cartoon beep beep

Generative adversarial network. Jargon fog for inputs along the way to an output

Generative AI. Reason to fire writers and PR people

Google Gemini. An example of tech bro-loney from an ad sales outfit

Guardrails. Stuff to minimize suicides, law suits, and the proliferation of chemical weapons

Hallucination. Errors

Inference. Guesses

LLM. Today’s version of LSMFT

Machine learning. Math from half century ago

Microsoft Bing. Beats the heck out of me

Multimodal AI. A fancy way to say words, sound, pix, and video to help un-employ humans or did this type of work

Natural language processing. Software that understands William Carlos Williams’ poetry

Neural network. Lots of probability and human-fiddled thresholds

Open weights. You can put your finger on the scale too

Overfitting. Baloney about hallucinations, being wrong, and helping kids commit de-living

Paperclips. Less sexy than The Terminator but loved by tech bros who like the 1999 film Office Space

Parameters. Where you put your finger on the scale to fiddle outputs

Perplexity. Another example of tech bro-loney

Prompt. A query

Prompt chaining. Related queries fed into the baloney machine

Prompt engineering. Hunting for words and phrases to output pornography, instructions for making poison gas, and ways to defraud elders online

Prompt injection. Pressing enter after prompt engineering

Quantization. Jargon to say, “We won’t need so much money now, Mr. Venture Bankman”

Slop. Outputs from smart software

Sora. Lights, camera, you’re fired. Cut.

Stochastic parrot. A bound phrase that allowed Google to give Timnit Gebru a chance to find her future elsewhere

Style transfer. You too can generate a sketch in the style of Max Ernst and a Batman comic book

Sycophancy. AI models emulate new hires at McKinsey & Company

Synthetic data. Hey, we just fabricate data. No copyright problems, right

Temperature. A fancy way to explain twiddling with parameters

Text-to-image generation. Artists. Who needs them?

Tokens. n-grams but to crypto dudes it’s value

Training data. Copyright protected information, personally identifiable information, and confidential inputs in any format, even the synthetic made up stuff

Transformer model. A reason for Google leadership to ask, “Why did we release this to open source?”

Turing test. Do you love me? Of course, the system does. Now read the sponsored content

Unsupervised learning. Automated theft of training data

Weak AI (narrow AI). A model trained on a bounded data set, not whatever the AI company can suck down

Zero-shot learning. A stepping stone to artificial intelligence able to do more than any miserable human

I love smart software.

Oh, the cited source leaves out OpenAI’s ChatGPT. This means “Titanic” after the iceberg.

Stephen E Arnold, January 5, 2025

Written by Stephen E. Arnold · Filed Under AI, News, Technology

Comments

Got something to say?

Search the site
Subscribe to Beyond Search
Feature archive
News archive

Stephen E. Arnold monitors search, content processing, text mining and related topics from his high-tech nerve center in rural Kentucky. He tries to winnow the goose feathers from the giblets. He works with colleagues worldwide to make this Web log useful to those who want to go "beyond search". Contact him at sa [at] arnoldit.com. His Web site with additional information about search is arnoldit.com.