AI Superpowers, the new world order

I thoroughly enjoyed Kai-Fu Lee’s book ‘AI Superpowers China, Silicon Valley, and the New World Order when I first read it in 2018. At the time, both finding the content immensly interesting yet also considering it still something that was some way down the road.

ChatGPT and the rise of the LLM

2022 and now 2023 has seen an explosion in LLM news, with Open AI’s ChatGPT probably taking most column inches followed closely by Google’s Bard. More recently the big Chinese tech companies have been unleasing their own LLMs as well.

My thought that Kai-Fu Lee’s description of future AI capabilities being some way down the road has arrived quicker than I had imagined it might.

What is an LLM?

As a quick note; the Large Language Model or LLM is a model that has been trained using vast quantities of unlablled text and self-supervised learing.

There is no formal definition for an LLM, but it is commonly understood to have parameter counts in the billions.

If Data is the new oil, will China be the new Saudi Arabia?

Kai-Fu Lee has a really interesting background, he’s seen the best of the US Tech World and the best of the Chinese Tech World. He has ideas as to where one migh have advantages over another, I think one of the really interesting points that stuck with me was that the Chinese tech scene is far more aggressive than the US, brute force, quantity over quality and ultimately less data protection regulations in China, allow their tech companies to push into spaces that US companies could never venture.

Ultimately he predicted China would likely win the artificial intelligence arms race.

China might be scared of what an AI LLM might say.

One thing that Kai-Fu Lee perhaps didn’t consider was that the Chinese regulators might not be too keen on LLM’s or at least ‘this stage’ of the AI roadmap.

Recent news suggests that China might actually want to limit the freedoms of LLMs, preventing them from offering answers in some areas for example. This may ultimately hold the chinese AI companies back when compared to their US competition.

Specifically, Chinese regulators are prohibiting things like any content that subverts state power, advocates the overthrow of the socialist system, incites splitting the country or undermines national unity.

So we’ll need to keep an eye on how this develops. I suspect that what we’ll end up with is a bit of middle-ware, something that can decide if the question should be allowed to asked and if so, pass it to the LLM, if not, either return an approved message or simply reject the question. This middleware approach would allow the LLM to develop at a competetive rate to those in ‘The West’ and perhaps offer a different answer for users outside of China, yet ensure that users inside the PRC are served with information that meets the countries regulatory requirements.

Job fears.

Kai-Fu Lee explores the potential impact of AI on the job market, predicting that while AI will eliminate many jobs, it will also create new ones and increase overall productivity.

Some of the communities that I’m a member of have been flooded with questions like ‘is this the end of my job’ or ‘what’s the point in learning how to do X’ and similar. People understand that this technology will change the world, but don’t really know how.

The sad reality is that a lot of jobs will go, but new jobs will replace them, my favourite quote being that It is not AI that will replace your job, it will be an AI operator. I think that an important middle step on the road to the new world order, will be understanding how to use the AI tools, how to ask the right questions, this will be the short-medium term challenge.

Kai-Fu Lee argues that education and retraining will be essential for workers to adapt to the changing job market and I agree with this, it certainly isn’t the time to throw the towel in and call it quits waiting for the government or the tech whales to pay us all to be users of, and used by, technology.

Time to re-evaluate the high-earning careers

Another point that stuck with me in Kai-Fu Lee’s book was that there should be a rebalancing of jobs, with Care Givers essentially being the jobs that an AI will never really be able to replace. Mostly due to a lack of ability to show empathy.

From this we might consider that in a typical hospital, it would not be the doctors that are at the top of the pay-scales any longer, as a doctor and a specialist are really just people that paid to study and have the ability to remember a lot of information.

If there is one thing we can realise about AI is that remembering the information part is covered. Therefore the real top earners should be the nurses, carers and similar. The people that today will likely be earning significantly less than a specialist in the same hospital.

Someone that can sit and share an elderly persons time, listen to and show enjoyment for their stories. Or someone that can make the comfort of a dying person their top priority. This is a role that currently could not be replaced by an AI. Not if we care about the receiver of the care anyway.

Future Predicted?

Kai-Fu Lee predicted a lot of what we’re seeing now. Though surprsingly, it may be Chinese regulation that slows down the AI cold-war more than perhaps he imagined.

China has started to aggressively regulate their tech industry in the years since this book came out for a number of reasons, some of which you can imagine for yourselves.

AI Regulations applied to the world outside of China

One thing I am interested to see is how Chinese tech companies might try to exploit international markets, there are things that may not be allowed to happen in China that could happen in other places. ‘The West’ can be accused of looking down on ‘The East’ because ‘they’ don’t share our values. Yet we’ve let the Web2 Social Media companies make immense sums of money on the back of the data that they’ve harvested from users of their systems. Regulation of this data should have been prioritised.

So just how clever could an LLM designed and built in China but unleashed on the West become?

Chinese firms working on LLM technology

One company that, for me at least, seems to have been caught on the back-foot, is ByteDance (makers of Douyin and TikTok) which from what I can tell has only just started building a team to create it’s own LLM. They clearly have some clever folks over there, so I’ll be interested to see what they churn out.

ALIBABA

Tongyi Qianwen will initially be integrated into DingTalk, Alibaba’s workplace messaging app and Tmall Genie, Alibaba’s voice assistant.

TENCENT

HunyuanAide early news suggests this is a ChatGPT clone that will be rushed out as soon as ready. The news is light, it may be just enough to keep Tencent in ‘the news’ for now, just enough to keep them in the conversation until they’ve got something to really talk about.

CHINA TELECOM

Developing an industrial version of ChatGPT for telecommunications, which will use AI in some customer service functions, again, is this just a chat bot or are we talking about something more?

BAIDU

Baidu recently unveiled its AI-powered chatbot known as Ernie Bot, currently available to a limited number of users who apply for access codes.

Baidu plans to use Ernie Bot to revolutionise its search engine, by far the most dominant in China, as well as use it to increase efficiency in its cloud, smart cars and household appliances businesses.

360 SECURITY TECHNOLOGY INC

Claims it has an LLM behind closed doors but that it could not give a clear indication on when it would launch any related products.

SENSETIME

April 10 unveiled a number of new AI powered products including a chatbot and image generator based off its AI model SenseNova.

KUAISHOU TECHNOLOGY

This is a TikTok type player. The LLM is at research stage.

FUDAN UNIVERSITY TEAM

MOSS another ChatGPT clone, possibly just something to try and stay relevant, it will be interesting to see what comes out after more time and investment.

JD.COM

ChatJD and will be aimed at serving other businesses, probably nothing more than a powerful chat bot based on early news.

NETEASE

Plans to deploy large language models technology to serve its education business, a source told Reuters.