#sema

#sema

Buyers no longer open ten tabs, skim through blog posts, and slowly form an opinion over weeks. Instead, they ask a single question to an AI system and receive a shortlist in return, usually two or three companies that feel familiar, credible, and safe enough to justify internally. That shortlist often becomes the entire market in the buyer's mind.

I'm one of those authors whose books AI ate for lunch a few years back. At some point I might get a check to pay me for a dozen years' work on the three books it stole, but really, there's no way to compensate for the fallout. AI seems to think no, it can't think, only shuffle what real people thought that a machine can write as well as a person can.

For the first time, speech has been decoupled from consequence. We now live alongside AI systems that converse knowledgeably and persuasively-deploying claims about the world, explanations, advice, encouragement, apologies, and promises-while bearing no vulnerability for what they say. Millions of people already rely on chatbots powered by large language models, and have integrated these synthetic interlocutors into their personal and professional lives. An LLM's words shape our beliefs, decisions, and actions, yet no speaker stands behind them.

Grammarly is now offering 'expert review' of your work by living and dead academics. Without anyone's explicit permission it's creating little LLMs based on their scraped work and using their names and reputation.

By comparing how AI models and humans map these words to numerical percentages, we uncovered significant gaps between humans and large language models. While the models do tend to agree with humans on extremes like 'impossible,' they diverge sharply on hedge words like 'maybe.' For example, a model might use the word 'likely' to represent an 80% probability, while a human reader assumes it means closer to 65%.

Semantic ablation is the algorithmic erosion of high-entropy information. Technically, it is not a "bug" but a structural byproduct of greedy decoding and RLHF (reinforcement learning from human feedback). During "refinement," the model gravitates toward the center of the Gaussian distribution, discarding "tail" data - the rare, precise, and complex tokens - to maximize statistical probability. Developers have exacerbated this through aggressive "safety" and "helpfulness" tuning, which deliberately penalizes unconventional linguistic friction.

A major difference between LLMs and LTMs is the type of data they're able to synthesize and use. LLMs use unstructured data-think text, social media posts, emails, etc. LTMs, on the other hand, can extract information or insights from structured data, which could be contained in tables, for instance. Since many enterprises rely on structured data, often contained in spreadsheets, to run their operations, LTMs could have an immediate use case for many organizations.

What happens under the hood? How is the search engine able to take that simple query, look for images in the billions, trillions of images that are available online? How is it able to find this one or similar photos from all that? Usually, there is an embedding model that is doing this work behind the hood.

Last year, a talented programmer friend of mine decided to give vibe coding a try. Vibe coding is the practice of describing to an AI chatbot what kind of program you want, and letting the AI write it for you. In a matter of minutes you can have new software in front of you, and just start using it. At least, in theory. This is what LLMs (Large Language Models) are supposed to be best at - generating usable software for professional developers