Yesterday’s Chinese company iFlytek Throwing itself into OpenAI’s bread and butter by announcing a product meant to compete with ChatGPTThe company’s “Spark Desk” has been described by the company’s founder and president, Liu Qingfeng, as a “cognitive big model” and even “the dawn of artificial general intelligence”. But beyond those buzzwords, there was also the promise that Spark Desk would surpass OpenAI’s ChatGPT by the end of the year.
Hopefully some of the above can be applied to corporate marketing buzzwords. If I had to write an article announcing that artificial general intelligence (AGI) is here, I can assure you my mind would be elsewhere. Perhaps even more so if that AGI is Chinese. I’m not sure you can trust AGI, which social scoring systems consider to be the “cognitive big model” bread and butter.
That aside, there are many interesting elements in this release. Every day we hear another ChatGPT spawn, whether it’s officially or unofficially linked to OpenAI’s work. Given the status quo of the technology’s impact (even if its impact is still opaque and largely unrealized), it stands to reason that all players with sufficient funding and expertise will pursue their own models. was. .
The question, of course, is whether iFlyTek and Spark Desk can actually live up to their claims, especially that they’re one ahead of OpenAI in their own games. The answer can vary depending on multiple factors and how you look at the subject.
ChatGPT is not intended for the Eastern public.training data, linguistic and cultural gaps (opens in new tab) This distinguishes ChatGPT’s impact on the East Coast compared to the western world. By its definition, “Spark Desk”, given sufficient maturity time, has the potential to offer a much improved (and better) user experience for Eastern users compared to ChatGPT. I have. Maybe by the end of the year it will happen. It certainly already offers a better experience, especially for Chinese users. Preemptively ban ChatGPT from traversing the Great Firewall (Except Hong Kong).
The decision to ban ChatGPT may have stifled innovation that would otherwise have been sparked. Just look to our own news outlets to see how many industries are affected by this technology. It is something no country can give up willingly on a whim. It was only a matter of time before a capable competitor was announced.
We’ll have to wait a few years to see if iFlytek’s claims come true or die. Comparing the two her LLMs quantitatively is difficult, especially when the target audience cultures are very different. One thing is certain, OpenAI is not just resting on its laurels and waiting for other industry players to catch up.
The ChatGPT version that iFlytek’s Spark Model has to deal with is different from the GPT we know today. Perhaps OpenAI’s expertise and time-to-market advantage will give it an edge over the competition (and that’s what we expected). But you should also remember that there are multiple ways to achieve the desired result. US technology sanctions against China have not been as effective as hoped, with China willing to bear the burden (and cost) of training cutting-edge technology on obsolete and superseded hardware. shown. Millions of dollars and hundreds of hours of extra training time are damned.
A few more billions of dollars might be enough to fill the gap. At least, that’s China’s bet.
