Article Complete
AI FeaturesMathletes Outmatched!

Inside The Mind of Conrad Wolfram

“Game over for humans on that one,” commented Wolfram on the math solving capability of ChatGPT + Wolfram plugin

Inside The Mind of Conrad Wolfram

Key takeaways

  • Leverage AI advancements by integrating ChatGPT with Wolfram for enhanced math problem-solving.
  • Recognize the significant improvement: ChatGPT scored 43% alone, but 96% with Wolfram's plugin.
  • Acknowledge the shift towards an AI-driven society as highlighted by Conrad Wolfram.
  • Be cautious of exaggerated claims about AI capabilities; balanced expectations are essential.
  • Embrace technological revolutions while remaining aware of potential challenges and problems.

Takeaways drafted with AI from the article text.

Since its release, the high profile ChatGPT has been infamous for its poor maths skills. The subject has indefinite problems of its own and the AI bot is hardly any good at providing solutions.  

To bridge the AI and maths gap, earlier this year, Wolfram Research joined hands with OpenAI to roll out a plugin. ”We have seen some interesting results with our LLM. I tried to run a British ‘A’ level maths, an exam students take before University, and ChatGPT alone got 43% which is quite impressive, but Wolfram plus ChatGPT got 96%,” Conrad Wolfram, strategic director, European co-founder and CEO of Wolfram Group, told AIM.  

“Game over for humans on that one”, Wolfram chuckled.

On the contrary, what annoys him the most is ‘experts overclaiming’ things. He advises, “Don't put everything in this particular basket.” He is optimistic that like all big technological revolutions, this will be positive, but there will be problems.” Furthermore, he is frustrated with the idea that 'we seem to have found the solution to everything'. “That's an impressive new technology but it doesn't solve everything,” he said. 

The Wolfram Dialogue

The introduction of GPT models has led to very interesting conversations across all the fields including education. It has brought new perspectives among the Wolframs, too. “One thing that I've realised recently with our computer-based math programme is that we need to educate the teachers, as well as try to find the right content for students. A few weeks ago, the unusual framework we built for mapping out a curriculum, is like the framework for teaching an LLM a particular subject.” 

Programming has been another topic of discussion recently, Wolfram revealed. “Programming is a skill that is absolutely sought after at the moment. It may have changed a little because of ChatGPT; there'll be less of humans writing the first cut of a programme, but I think programmes will still be there. There will be more editors of code than makers,” he opined. 

What Language would You Want your LLM to Write in?

The maths genius thinks it is important to have an abstract language. Some people believe you don't need that anymore. Disagreeing with them, Conrad said, “I think you need the precision of an abstract language for many things. You can’t always just talk to an LLM in English and hope that it will do the right thing. You do need computer languages”.

Plugging in the famous Wolfram language he said, “The problem from a human point of view is there's a lot of vocabulary to learn. There are about 6500 indigenous Wolfram language functions. So people find that daunting and they always think that they haven't learned in a sense the right language, the right function.” He further praised the capability of LLMs as “fantastic because it has no problem at all knowing those functions and piecing them together as a first cut of the programme”. 

An Educational Shift

“As we talk about education, there’s no sudden thing that causes change,” believes Wolfram. 

He recalled a conversation with someone who had previously been a minister of defence, and then a minister of education. In defence, he told Wolfram, you suddenly get attacked, everything has to change to defend yourself and that pushes everything forwards. The trouble in education is, it sits there with nothing to cause a sudden transformation. “Now we've got two things that potentially caused a transformation -- the pandemic and ChatGPT,” he said.  

Proposing ways to technologically advance the maths curriculum, Wolfram said, “One is how we improve the pedagogy and personalise it better for all students in all subjects. The second is focussing on just the right subjects to teach because the world has changed. He suggested, “Start with real problems”. 

One of the problems Wolfram often presents when giving talks is, "if we close all the rooms down, all the air ventilation down, how long can the people in the room survive?" That's a problem about how much air there is in the room. It can be turned into an equation to get some value for the time. 

“Mathematics is fantastic for understanding how problems solve and why you can't just talk about it in your language. I would start from problems in the real world to which we want to apply mathematics to get the decisions,” he concluded.

Published

Corrections: AIM corrects errors of fact and notes the change on the story. Spotted one? Tell the editors. Read our editorial standards.