>> The most obvious answer is AI, because ChatGPT is extremely useful as a coding companion. People keep saying that but every systematic test of LLM coding ab

>> The most obvious answer is AI, because ChatGPT is extremely useful as a coding companion. People keep saying that but every systematic test of LLM coding ability suggests otherwise. For example, another article linked to this table of GPT-4 results on various tests yesterday: https://www.thenewatlantis.com/wp-content/uploads/2023/07/Schulman-image-1.png https://www.thenewatlantis.com/wp-content/uploads/2023/07/Sc... The table is from OpenAI's article on GPT-4: https://openai.com/research/gpt-4 https://openai.com/research/gpt-4 And it shows how well GPT-4 did on verious tests, like LSAT, etc. One of the tests is listed as "Codeforces Rating". The score is: 392 below 5th [percentile]