(Image credit: iFlytek)
By unveiling a product yesterday that is intended to compete with ChatGPT, the Chinese startup iFlytek launched a direct assault on OpenAI’s core business. Liu Qingfeng, the company’s creator and president, referred to the “Spark Desk” as a “cognitive big model” and even as the “dawn of artificial general intelligence.” However, there was a promise that went beyond these catchphrases: by the end of the year, Spark Desk will outperform OpenAI’s ChatGPT.
We should be glad that some of the aforementioned are only buzzwords used in corporate marketing. If/when I have to publish an essay stating that Artificial General Intelligence (AGI) is here, you can be sure that my thoughts won’t be on that. Perhaps much more so if the AGI was Chinese because I’m not sure I can trust one that believes social scoring systems are the lifeblood of its “cognitive big model.”
Leaving that aside, this edition has a number of intriguing features. Every day, whether formally or informally connected to OpenAI’s efforts, we learn about a new ChatGPT spawn. Given the current state of the technology’s influence (even if that impact is still hazy and largely unrealized), it was only inevitable that each participant with the resources and know-how to pursue their own models, tailored to their particular public and stakeholder base, would do so.
Of course, the real question is whether iFlyTek and Spark Desk can live up to their promises, particularly the assertion that they can outperform OpenAI in its own game. The answer will probably rely on a number of variables and how you feel about the issue.
The Eastern populace wasn’t intended for by ChatGPT. A training data, language, and cultural divide exists between the influence of ChatGPT on the Eastern seaboard and the Western hemisphere. And by that criteria, it’s perfectly likely that, given enough development time, “Spark Desk” will provide Eastern users with a significantly better (and more relevant) user experience compared to ChatGPT. Maybe that will happen even before the year is over. For Chinese users in particular, it already provides a better experience due to the nation’s preemptive restriction on ChatGPT from going through its Great Firewall (apart from in Hong Kong).
Banning ChatGPT probably prevented innovation that it would have otherwise stimulated. We only need to look at our own news sources to understand how many different industries technology is affecting. It’s impossible for a nation to voluntarily abandon that, and it was only a matter of time before a worthy rival emerged.
We won’t know whether iFlytek’s assertions are true or false until the end of the year. Comparing the two LLMs quantitatively will be challenging enough, especially given how different their target audiences are from one another culturally. One thing is certain: OpenAI won’t just sit back and wait for other participants in the market to catch up.
The ChatGPT version that iFlytek’s Spark Model must deal with won’t be the same GPT that we are accustomed to. Perhaps OpenAI’s experience and advantages in time to market will maintain it in the lead in the race (and that’s what we’d anticipate); however, we also need to bear in mind that there are several ways to accomplish a desired goal.
Also read : TikTok ban by the EU, according to China, undermines business confidence
It has been established that the U.S.’s technological sanctions against China have not had the desired impact and that the nation is willing to bear the burden (and costs) of paying for the training of cutting-edge technology in obsolete, superseded hardware, regardless of the millions of dollars and additional training hours.
It might just take a few extra billions to close the gap. China, at least, is betting on that.