OpenAI Accuses DeepSeek of Copying Its Copy-Detection Technology
OpenAI and Microsoft are feeling a bit salty over Chinese AI startup DeepSeek's meteoric rise and potential poaching of some of their tech secrets. It's a curious claim coming from the creators of ChatGPT, a program they once acknowledged couldn't exist without the free usage of all copyrighted data worldwide.
OpenAI brought their concerns to The Financial Times, claiming they had proof DeepSeek employed a method known as "distillation" to develop their successful chatbot. I'll let tech guru and White House AI and Crypto czar, David Sacks, explain.
"There's a technique in AI called distillation," Sacks explained to Fox News. "It's when one model learns from another model. The student model poses questions to the parent model, just like a human would learn. AI can do this asking millions of questions, ultimately mimicking the reasoning process it learns from the parent model."
In simple terms, distillation is like one AI learning from another through a series of questions, essentially absorbing its knowledge.
OpenAI suggests DeepSeek obtained outputs from ChatGPT and utilized these to train their own bot, effectively creating a Language Model (LLM) at a fraction of the billions spend on training ChatGPT. But OpenAI declined to share evidence supporting this claim, stating that no company can directly "copy" ChatGPT.
Sacks, speaking to Fox News, hinted at the possibility of cramping DeepSeek's style:
"There’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI’s models," Sacks said, without providing said evidence. "I don’t think OpenAI is very happy about this. I think one of the things you’re going to see out of this is our leading AI companies taking steps to prevent distillation."
Bloomberg dug deeper, citing a source within OpenAI who outlined a ruse: Microsoft had identified suspicious activity around ChatGPT's API and flagged the company. The API calls could be linked to DeepSeek's distillation of ChatGPT. This might infringe upon ChatGPT's terms of service.
DeepSeek dethroning ChatGPT on Apple's AppStore happened just as AI companies saw hundreds of billions in market value evaporate from the stock market earlier this week. Income-generator NVIDIA, a manufacturer of the chips used in training chatbots, took a hit exceeding $500 billion.
Sacks offered reassurances to Fox News viewers, reiterating America's AI dominance:
"There are still great advantages to having an enormous number of chips. This is an area where America could continue to lead, thanks to a build-out of this infrastructure and having the most advanced chips," Sacks said. "So I think it's a little bit of an overreaction to say that America does not need AI data centers anymore."
Sacks pointed to two reasons China was gaining ground in the AI race: strict regulations imposed on AI companies by the Biden administration and AI companies becoming too focused on social justice issues.
"I think that our AI companies got a little distracted...I think that maybe they got a little bit complacent. They didn’t realize how close these Chinese companies were to them. They wasted a lot of time on things like DEI. You saw there was woke AI," Sacks said. "The models were basically producing things like black George Washington."
As of now, DeepSeek remains the reigning champ on the App Store.
Insight: Distillation is a method to create a smaller AI model capable of mimicking the capabilities of a larger, pre-trained model by utilizing its outputs. When supposedly used without authorization, it can violate the terms of service and raise significant legal and ethical concerns.
The future of AI technology could see companies taking measures to prevent distillation, as demonstrated by Sacks' hint at preventing DeepSeek's alleged use of this method. Artificial Intelligence, especially in the chatbot sphere, relies heavily on substantial investments for training, making it an attractive target for distillation.