>> For the first time in all of history, we can create functions in programming by training them, rather than writing them by hand. In other, slightly more tech

>> For the first time in all of history, we can create functions in programming by training them, rather than writing them by hand. In other, slightly more technical words, we can infer a function from observations and then use it. Like babies and toddlers do. This is historic, it’s transformative and it is a fundamental breakthrough. Nothing could be further from the truth. Program Synthesis [1] has been an active field of research with many real-world applications since the dawn of computer science. Even program synthesis from natural language specifications is not new [2]. Neural program synthesis is certainly not new [3]. Program synthesis from examples of program behaviour, without examples of programs (i.e. what Large Language Models can't do) is also nothing new [4]. None of that is new. None of that is even particularly well done by Large Language Models (what the author means when they say "AI" in the article). The author is misinformed. The rest of the article is also rife with misunderstandings and fautly assumptions but Program Synthesis is one of my areas of knowledge so that bit stuck out as a sore thumb. Please try to be better informed before writing lengthy articles to share your thoughs on the internet. _______________ [1] https://people.csail.mit.edu/asolar/SynthesisCourse/index.htm https://people.csail.mit.edu/asolar/SynthesisCourse/index.ht... [2] For example: https://www.microsoft.com/en-us/research/wp-content/uploads/2016/07/NLPSynthesis.pdf https://www.microsoft.com/en-us/research/wp-content/uploads/... [3] For example: https://sunblaze-ucb.github.io/program-synthesis/index.html https://sunblaze-ucb.github.io/program-synthesis/index.html [4] https://link.springer.com/referenceworkentry/10.1007/978-0-387-30164-8_399 https://link.springer.com/referenceworkentry/10.1007/978-0-3...