AI · អត្ថបទ
Stop Counting AI Tokens and Start Measuring What Actually Gets Shipped
MIT Technology Review ·
មើលត្រួសៗ
AI agents call models many times for a single coding task, so token costs (the smallest units AI uses to read and write text, which determine billing) can climb quickly. This article argues that companies should judge AI spending by the cost of each merged code change, not by total usage.
ចំណុចសំខាន់ៗ
- A single agent task can use 10 to 50 times more tokens than a simple chat.
- Some companies used up their AI budgets early in the year, and GitHub Copilot switched from a flat fee to a usage-credit model.
- The article recommends running easy tasks on open-weight models (AI models whose weights are public and can be run directly) and using expensive top-performing models only for difficult planning.
- Running your own servers lowers the cost per call as usage grows, but electricity and operating costs must also be considered.
ហេតុអ្វីវាសំខាន់
If companies only try to cut token fees, they may block important development work, so it matters to measure costs against actual results.
ប្រភពដើម
Stop counting tokens, start counting what ships
MIT Technology Review · Nov 2, 22:10 KST
Read on MIT Technology Review ↗AI summarized the original article. Check the original for accuracy. Copyright belongs to MIT Technology Review.
បន្ថែមពីវិស័យនេះ
- OpenAI Opens ChatGPT Space to Plus Users for Pages, Slides, and Team Sharing
OpenAI opened ChatGPT Space to Plus users on the 20th, local time. Space works as a new home for pages, slides, images, files, and uploads. Users can write pages with ChatGPT, add interactive charts, turn ideas into slides, and invite others to edit together.
AI 매터스 · Nov 21, 11:26 KST - Anthropic Researcher Claims Claude Helped Prove Riemann Hypothesis, Experts Await Review
Wes Roth covers a claim by Anthropic researcher Levent Alpöge that he proved the Riemann Hypothesis with help from the AI model Claude. He looks at the huge Lean formal proof, reactions from mathematicians and AI safety researchers, and what this could mean for AI and math.
Wes Roth · Nov 21, 11:24 KST - AI Claims Proof of 167-Year-Old Riemann Hypothesis With Anthropic's Claude
An Anthropic researcher, Levent Alpöge, says he and Claude found a proof of the Riemann Hypothesis, a famous 167-year-old math problem. He released a version written in Lean, a language that checks math proofs by computer, and also claims to have solved a broader generalized version. Experts called it a potentially historic moment but said the proof still needs careful checking.
매일경제 · Nov 21, 11:21 KST - Korea Orders Overhaul of National Exams and Hiring to Stop AI Cheating
President Lee Jae-myung ordered a full overhaul of national exams and private company hiring to stop people from cheating with AI, from the college entrance exam to job exams. A government-wide task force and a Blue House task force were launched at the same time.
서울경제 · Nov 21, 10:35 KST