Have we entered the age of AI?
2023-02-12

AI has been a concept for decades. In the 1950’s renowned computer scientist Alan Turing proposed the imitation game or Turing test. Could a computer pass itself as human? Skip ahead to the present day, and we have Open AI publicly releasing the language model Chat-GPT. Having used this model it is clear that it would be very possible to have a human believe they were conversing with another human when presented with Chat GPT’s answers. However the safety restrictions put in place by OpenAI mean it would be hard to undertake a proper turing test as it is quite easy to ask a question which leads to a restricted answer. Google has a similar language model called LaMDA, though this is not at the time of writing publicly available. One of their employees was dismissed due to having convinced themselves that the language model had gained sentience. Despite the way these models are generated making it impossible for the AI to have any “thoughts or decision making”, being able to trick us into believing it can, is still a very powerful and important milestone in the history of AI.

The AI race is on with Microsoft (in partnership with Open AI) and Google investing billions of dollars into developing their intelligence capabilities. Currently the bulk of these language models are being used to enhance search engines, but I imagine soon we will see them perform first line tech support, and assisting professions with interpreting large amounts of textual data, such as lawyers and doctors. As the API’s for these models become more available I think these models could end up impacting a vast number areas in a way not seen since the introduction of the internet.

Like the internet, AI propagation will bring risks and danger. Language models are not the only area of AI research that has made leaps and bounds recently. Researchers at Microsoft have produced VALL-E, a system which can mimic a human voice based on short snippets of the original voice. Open AI has DALL-E which can turn a textual prompt into an image. It is not hard to imagine that in the near future, if not today, we will no longer be able to tell what is real and what has been artificially generated. In Hollywood the fear is that AI will overthrow humanity, although we are a long way from that reality the current reality is just as chilling. Unchecked AI and an uneducated population could lead to a world where evidence becomes unreliable, fraud becomes widespread and the truth becomes questionable.