[Elon Musk Shares Self-Mocking AI Meme: From Proving Theories to Just Stacking GPUs] According to monitoring by Dongcha Beating, Elon Musk shared a comparison meme titled 'AI then / AI now,' commenting with just one word: 'Accurate ?'. The image illustrates how, in the past, machine learning research involved designing kernel methods, proving why models were effective, and building probabilistic models. In the era of large models, the process has been reduced to three steps: make the Transformer bigger, feed it more data, and add more GPUs—then the model becomes stronger yet again. This highlights the current dilemma in the large model industry: in the past, principles were studied first, and then models were designed; now, scaling up often comes first, with research into why capabilities emerge coming later. [Original article link]