As a tech journalist for the last 20 years, I’ve had a front-row seat to the slow death of the English language, driven by the engineers and marketers of Silicon Valley who use clunky abbreviations, awkward jargon, and meaningless superlatives to describe the latest innovations.
Let’s chat about U.G.C.! Did you know Google’s assistant can now have a two-way conversation? This new smartphone is shockingly, if not stunningly, faster than the last one.
Lately, the relentless verbiage has worsened, with more superfluous adjectives, buzzwords, acronyms, and abbreviations— such as superintelligence, RAG, and TPU—added to the list at an alarming pace.
The obvious culprit is the artificial intelligence boom that has upended the tech industry, creating a new lexicon. Notably, the dictionary publisher Merriam-Webster chose “slop” as its word of the year, referring to the AI-generated content that polluted our social media feeds.
Tech companies like Nvidia and Dell have named their newest data centers “AI factories.” Companies describe themselves as specialized data centers that require large amounts of storage and power to operate AI systems. Historically, tech giants have upgraded large-scale computing facilities to support new technologies, so AI factories are, simply put, data centers.
U.G.C.
No, it’s not a shoe brand. It stands for user-generated content, and the abbreviation has been popular lately among Google employees working on A.I. search technology. In plain terms, they are referring to social media posts, such as a TikToker talking about their favourite burger restaurant.
A.G.I.
For years, companies including OpenAI, Google, and Amazon have said their goal is to achieve A.G.I., meaning artificial general intelligence, a technology with humanlike cognition. But for decades, “artificial intelligence” has referred to technology that mimics the human brain. The elusive nature of A.I. raises questions about whether other tech products labeled “A.I.” are actually intelligent.
Superintelligence
Even though it’s unclear when, if ever, the tech industry will achieve A.G.I., Mark Zuckerberg of Meta is already talking about the next phase. When AI technology becomes powerful enough to capture data about everything we see and hear in real time, humanity will achieve superintelligence, he predicts. (In September, when Mr. Zuckerberg publicly demonstrated a pair of computerized glasses that could one day deliver superintelligence, Meta’s AI got stumped when asked how to make a steak sauce.)
RAG
This acronym stands for retrieval-augmented generation, a technique used to improve chatbot accuracy. It involves connecting a chatbot to external information sources, such as an encyclopedia, a history book, or a news article. It’s an off-putting acronym, but think of it as a rag that can clean up occasionally messy answers spewed by chatbots.
Multimodal
This tongue-twister of a word describes technology that can answer your questions about images, text, and audio files you share with a chatbot like ChatGPT or Gemini. You will hear this term more often in the coming years as companies release smart glasses with cameras and microphones, enabling an AI assistant to provide information about what you see and hear.
NPU
Most consumers probably wouldn’t care whether a computer shipped with a neural processing unit, a chip that accelerates AI apps that generate text and images. Nonetheless, Microsoft, Dell, and Lenovo are highlighting NPU chips to market their newest laptops. The chips are simply faster and more energy-efficient, as new computer chips tend to be.
Related: TPU, or tensor processing unit, a term that Google uses to describe the neural processors it relies on in data centers to make AI software work.
Vibecoding
Chatbots like Claude and Gemini can automatically generate code, enabling inexperienced programmers to write simple programs by typing a prompt such as “I want to create an app to choose an outfit from my closet.” Enthusiasts have called the ritual “vibecoding,” and the results have been hit or miss.
Agentic
When a chatbot does something for you, like booking a flight, techies call this “agentic,” referring to how chatbots can act as agents, similar to the people who book your travel. The clunky word has gained traction in the last few years, but “virtual assistant,” the term used to describe older tools that tried to help you (e.g., Siri and Alexa), was less cringe.
Magic
When Steve Jobs introduced the first iPhone in 2007, he said the touchscreen “works like magic,” referring to its performance. This year, Google awkwardly used the word for a new AI tool it released for smartphones, Magic Cue.
But Magic Cue and similar AI technologies that automate tasks require you to share large amounts of personal data, such as your contact list, location, messages, and email. There’s nothing magical about that at all.
