Relevant and irrelevant fears of Artificial Intelligence

By Chandre Dharmawardana

The oracle-like power of the ‘large-language’ Chatbot named has frightened rational techies and mystic mullahs alike. Elon Musk, Steve Wozniac who co-founded Inc., historians like Yuval Harari, and academics like the Turing-prize winner Yoshua Bengio of Montreal University called for a for developing AI beyond GPT-4, the latest technology released by OpenAI.

Much of the is available in the public domain and will be grit to would-be malicious urban terrorists. So, a six-month pause can little do once the horse has bolted from the stable? Just as with nuclear weapons, a dedicated team of engineers can, given sufficient time, develop its own code to serve a questionable master. So, there was really no way of closing the stable in time!

The use of OpenAI’s ChatGPT has gone viral, becoming learned sidekick of student, entrepreneur, army general, saboteur and spy, while secretly storing all it hears. But ChatGPT can and display psychotic behaviour. Will a GPT-4 App controlled by a surveillance state rule our neuron circuits, with human society behaving like an ant colony?

Such fantasies ignore the enormous computing power, needed to execute any large dystopian plans. Hence, demands to stop AI research on such fear is irrelevant and baseless.

The immediate danger of AI resides in its use by urban terrorists, and in its power to imitate and misinform, hurting and possibly holding to ransom individuals or organisations. It can create realistic images of people in compromising positions and blackmail them or use them to advance questionable political and commercial agendas.

Can the chatbot be used to discern political trends more effectively than campaign planners. The current chatbot has been trained on information up to 2021.

Here are the Chatbot answers to some questions relevant to how the Sri Lankan leadership, and its bid to resolve its economic crisis, are resolved.

Question:

Will Sri Lanka succeed in mitigating its foreign debt and will it get adequate help from the IMF?

Answer from ChatGPT:

“Sri Lanka’s ability to mitigate its foreign debt will depend on a number of factors, including its ability to implement effective economic policies, attract foreign investment, and manage its debt. The country has been facing challenges related to high levels of external debt, which have contributed to a decline in foreign reserves and a weakening of the currency.

The International Monetary Fund (IMF) has been working with Sri Lanka to help address its economic challenges, including providing financial assistance and technical support. However, the IMF’s support will depend on the country’s ability to implement necessary policy reforms and meet certain economic targets.

In order to mitigate its foreign debt, Sri Lanka may need to implement structural reforms, such as reducing government spending, improving tax collection, and promoting exports. The government may also need to seek assistance from other sources, such as bilateral lenders or multilateral development banks”.

Question:

What is your assessment of Ranil Wickremasinghe as a potential leader of Sri lanka in the context of its current crisis?

Answer from ChatGPT:

“Ranil Wickremasinghe is a veteran politician who has served as the Prime Minister of Sri Lanka multiple times. He has extensive experience in governance and has implemented several economic and social reforms during his political career.

In the context of Sri Lanka’s current crisis, Wickremasinghe’s economic policies and track record may be seen as potential assets, given his focus on trade liberalisation, attracting foreign investment, and improving economic growth. He has also shown an interest in promoting reconciliation among Sri Lanka’s diverse communities.

However, it is important to note that Wickremasinghe’s leadership style and policies have also been subject to criticism, particularly in relation to issues such as corruption, human rights, and minority rights. Additionally, his leadership of the United National Party (UNP) has faced some internal divisions and challenges.

Ultimately, whether Wickremasinghe would be viewed as a potential leader in Sri Lanka’s current crisis would depend on a range of factors, including his ability to build a broad-based coalition of support, his plans for addressing the country’s economic challenges, and his ability to address the concerns of diverse communities in the country.

Overall, the success of Sri Lanka’s efforts to mitigate its foreign debt will depend on a range of factors, including the government’s commitment to implementing necessary reforms, the support it receives from international organisations, and the broader global economic climate”.

So, we see that the chatbot will answer as ambiguously as any “political commentator” and we cannot expect chatbots to predict the stock market or political trends, although chatbots may do better than the “chattering classes” in reflecting the character of the news fed to it.

Those answers are based on the “training” received by the chatbot containing data up to 2021. Can chatbots help in innovative thinking, if they can be given continued training “on the job”? I believe the answer is a definite “yes”, although it is not yet clear if it can do the type of “reductionist” thinking that is needed in theoretical physics.

It was in 1955 that Herbert Simon and Allen Newell produced from Russell and Whitehead’s “Principia Mathematica” using AI to manipulate the limited language of symbolic logic. Today’s new chatbots trained on real languages spit out intelligent answers, but anchored within the orthodoxy of languages used to train the chatbot.

Such “creativity” is different to that of Newton who linked the fall of an apple with the “fall of a planet” towards the sun in moving in its orbit instead of moving in a straight line! Can a chatbot create that script which was NOT in any language in Newton’s time?

Basically, involves “fitting” thousands of non-linear parameters via the nodes and connections of an electronic analogue of a neural net and a memory. But the method of a Newton or Einstein is the very opposite of neural nets.

The physicist modeling presents a reductionist model with a bare minimum of parameters. In contrast, AI is mired in complexity itself; it is the methodological antithesis of theoretical physics. AI may provide “answers” to some of our questions, mimic and imitate some iconic figure but as yet provide no unifying theories.

When Goethe wrote, “Here I sit/ forming men/ In my own Image/ A race to be like me/…/And to defy you/As I do/, he was writing about form the Gods. Artificial Intelligence is a corollary of that Promethean act itself. There is no way to contain the horse always locked in the stable.

However, the more human-like the chatbot is, and the more it exceeds humans by its complexity and capacity, the more subject to black-swan events it will be. Its behaviour is entirely deterministic but . It has become a gun or a gismo of unknown capability and not subject to design.

So, the AI machines optimal for humans would be those with predictable behaviour and lesser complexity. Beyond that red line, even the AI designer, even if he/she were a super intelligent AI brain, is like a blind farmer unaware of what his harvest is going to be!

[Chandre Dharmawardana, is a scientist and author of “A Physicist’s View Of Matter And Mind” affiliated with Université de Montréal and the Quantum Physics group of the National Research Council of Canada.]