This could be adapted into a movie: the intern who sabotaged Bytedance (TikTok)’s AI, sued by ByteDance for 8 Million, wins Best Paper awards at NeurIPS 2024.
The full coverage of the previous drama can be found here. Here’s a reprise:
An intern at ByteDance, resentful over perceived unfair resource distribution within the company, launched a covert operation to disrupt the training of machine learning models. This technical team experienced significant setbacks due to the intern’s malicious actions, which were aimed directly at the core of their model training tasks.
The full account of these events was documented on GitHub (quite geeky indeed—you can check it out here). The intern’s tactics included altering PyTorch source code, randomly terminating multiple machine learning experiments, and injecting malicious code that interfered continuously based on colleagues’ troubleshooting efforts.
Btw, the intern himself (TianKeyu, according to FengHuang Net’s exclusive news report) has pretty impressive publication history and educational background.
And TianKeyu’s first author paper: Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction (see the OpenReview page here) won best paper of NeurIPS 2024. Btw, it’s said that he did this during his intern at ByteDance…
It’s good to see that the reviewers and the award committee aren’t biased by the negative legal issue surrounding an academic work—or, is there a possibility the negative new’s popularity breeds extra attention.
Who knows, time would tell.