This post is part 2 of EkaLore’s “What Can’t It Do” series concerning ChatGPT and other LLMs. Today’s post starts with a bit of history.
Alan Turing, one of the early giants of computer science, devised a test for Artificial Intelligence. An artificial device would have achieved intelligence if a human could not determine whether they were talking to a person or a machine. From the perspective of the 1940s, the amount of compute power necessary was unimaginable.
75 years later, we do have those unimagined levels of compute power. Today, the software can convincingly ‘converse’ with humans. The conversations are now to the point where a ‘software update’ caused an uproar when some individuals felt their ‘emotional connection’ to a computer program had been broken. Large flows of money had gone to the creators of the unreal personalities in the computers.
Software is capable of synthesizing ‘Chat’ and ‘images’ real enough to allow some people to form emotional attachments (leaving deeper discussion to others). The ‘AI’s ability to ‘listen’ to audible speech, ‘talk’ in a synthetic voice, chat in an engaging way, and ‘learn’ (able to form new responses based on history). The AI can also create ‘images’ synthetically convincing to the imagination. The software technology, and devices, have caught up to human expectations and perception of a level of dialogue sufficient to be emotionally engaging.
These technologies are exciting, but what good are they? Can they replace humans in commercial settings (sometimes)? Are they as good as humans? (or better in certain commercial circumstances.) Though people can form attachments to this new generation of chatbots, let’s not forget that people also formed attachments to Tamagotchi toys of the 1990s.
The next post in the series will talk a little more about the value of LLM powered Chat Bots for customer service and/or sales.
You can find this and other BAD (Big Data/AI/Data Science) posts at www.ekalore.com/bad-project-blog
Comments