Should You Threaten Your AI for Better Results?
·1157 words·6 mins
Sergey Brin claimed threatening AI models improves their performance. Andrej Karpathy echoed it. But when researchers ran nearly 400,000 tests, the results told a very different story.