Artificial Intelligence: ChatGPT and Others

ChatGPT (Generative Pre-trained Transformer) uses deep learning algorithms to analyse and generate text. It is a language model created with the purpose of holding a conversation with the end user. that allows human-like conversations. The model can answer questions, assist with tasks such as composing email, essays and code. It was launched in November 2022 by the artificial intelligence lab Open AI. 

What is the difference between ChatGPT and a search engine?

ChatGPT is a language model created with the purpose of holding a conversation with the end user. A search engine indexes web pages on the internet to help the user find the information they asked for. ChatGPT does not have the ability to search the internet for information. It uses the information it learned from training data to generate a response, which leaves room for error. 

This gives it a pretty wide range of abilities, everything from writing poems about sentient farts and cliché rom-coms in alternate universes, through to explaining quantum mechanics in simple terms or writing full-length research papers and articles.

Where we could spend hours researching, understanding and writing an article on quantum mechanics, ChatGPT can produce a well-written alternative in seconds.

The model was trained using text databases from the internet. This included 570GB of data obtained from books, webtexts, Wikipedia, articles and other pieces of writing on the internet. To be even more exact, 300 billion words were fed into the system.

Because the system is trained largely using words from the internet, it can pick up on the internet’s biases, stereotypes and general opinions. That means you’ll occasionally find jokes or stereotypes about certain groups or political figures depending on what you ask it.

DeepSeek is the name of a free AI-powered chatbot, which looks, feels and works very much like ChatGPT. That means it's used for many of the same tasks, though exactly how well it works compared to its rivals is up for debate. It is reportedly as powerful as OpenAI's o1 model - released at the end of last year - in tasks including mathematics and coding. Like o1, R1 is a "reasoning" model. These models produce responses incrementally, simulating a process similar to how humans reason through problems or ideas. It uses less memory than its rivals, ultimately reducing the cost to perform tasks. Like many other Chinese AI models - Baidu's Ernie or Doubao by ByteDance - DeepSeek is trained to avoid politically sensitive questions. 

When the BBC asked the app what happened at Tiananmen Square on 4 June 1989, DeepSeek did not give any details about the massacre, a taboo topic in China. 

It replied: "I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses."

Chinese government censorship is a huge challenge for its AI aspirations internationally. But DeepSeek's base model appears to have been trained via accurate sources while introducing a layer of censorship or withholding certain information via an additional safeguarding layer. Deepseek says it has been able to do this cheaply - researchers behind it claim it cost $6m (£4.8m) to train, a fraction of the "over $100m" alluded to by OpenAI boss Sam Altman when discussing GPT-4. DeepSeek's founder reportedly built up a store of Nvidia A100 chips, which have been banned from export to China since September 2022. Some experts believe this collection - which some estimates put at 50,000 - led him to build such a powerful AI model, by pairing these chips with cheaper, less sophisticated ones.

Excepts from article by Kelly Ng, Brandon Drenon, Tom Gerken and Marc Cieslak of BBC News January 28, 2025