Would Baidu’s answer to ChatGPT make a difference? • TechCrunch

Baidu, China’s largest search engine provider and robo-taxis developer, appears to be developing its own service for ChatGPT. The news, first reported by Bloomberg and The Wall Street Journal, sent Baidu shares higher on Monday to their highest level since September.

A Baidu spokeswoman declined to comment on the report. But it should come as no surprise that his Baidu, a self-proclaimed pioneer in China’s artificial intelligence field, is working hard to build a Chinese equivalent of today’s most powerful chatbots. The question is how much difference a tool can make, and where the limits are.

The driving force shaping China’s technological development over the past few years has been the rise of digital sovereignty. Digital sovereignty refers to the ability to control one’s own country’s “digital destiny”, allowing the AI ​​supply to include autonomy in the critical software and hardware of his chain. The episode of US export bans to China has prompted the Chinese government to demand more technology independence in areas ranging from semiconductors to basic AI research.

As OpenAI’s ChatGPT shows its potential to disrupt sectors ranging from education and news to the service industry, China’s tech leaders and policy makers are raving about how AI can be used to boost productivity at home. You may be pondering. China wants his ChatGPT, homegrown, not only to ensure control over the flow of data through such tools, but also to an AI that better understands local culture and politics. It is also for creating products.

Baidu’s conversational robot, due to debut in March, will first be integrated into the company’s search engine, according to The Wall Street Journal. This suggests that the chatbot produces results primarily in Chinese. Nevertheless, deep learning models are trained on both Chinese and English data sources, including the collected information. outside The Great Firewall, the country’s sophisticated internet censorship infrastructure.

That’s the interesting part. Like all other information channels in China, the Baidu chatbot is definitely subject to local regulations and censorship rules. As I wrote before, the company’s text-to-image application, his ERNIE-VilG, already rejects politically sensitive prompts. But conversational AI handles far more complex queries than image generators. How will Baidu walk the line between censorship imprisonment and leaving bots plenty of freedom and creativity?

Also important to machine learning performance are the underlying algorithms. According to The Wall Street Journal, Baidu adopted the “core breakthrough” Google developed and open-sourced in 2017, the algorithm that also inspired ChatGPT. However, there are most likely other significant pieces of proprietary algorithms that Baidu acquired or developed to form the backbone of its chatbots.

Hardware plays another important role in training large-scale neural networks. US chip sanctions against China pose a threat to China’s AI industry as companies lose access to the advanced semiconductors that power supercomputers and large data centers.

However, Baidu believes the chip ban will have “limited” impact on AI businesses, as we reported. In the short term, the company said, “We already have enough inventory. [chips] in hand. As for the future, Baidu will rely on its self-developed Kunlun AI chip to drive high-performance computing. Alternatively, you can increase the efficiency of your algorithm and reduce the work on your hardware.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *