1. Don’t Trust It
Even though AI has progressed immensely in the last couple of years, we’re not yet at the point where these models are infallible. There is no substitute for good old, human, common sense.
When an AI model produces a result that raises a brow, try asking yourself why it would deem the result as truth. Specifically, generative models such as ChatGPT and its predecessors in the GPT family have the tendency to “hallucinate” things – and sound awfully convincing while making up wild things.
AI models have even been developed to fact check these models:
2. Be wary of “Experts”
The field of AI is an ever-evolving field. By its definition, practitioners of the field attempt to simulate something that we do not yet understand fully – human intelligence.
Anyone who claims to know everything about this field clearly does not understand the nature of the field. The prevalence of these self-proclaimed experts have been further exacerbated by the availability of pre-built, no-code and auto-ML solution offerings.
While these solutions are incredibly useful and serve to democratise AI, they are meant to be used to prove concepts and not as production-grade implementations.
Some of the drawbacks of these no-code solutions are outlined here.
3. Don’t be afraid to take risks
As we have mentioned, the field of AI is growing and developing at a famously rapid rate. This means that we don’t yet know what the full set of use-cases look like.
A model capable of understanding natural language can easily be believed to be useful as a chatbot, but what about a personal legal expert? A paraphraser or copyrighter?
The field of AI, and Natural Language Processing (NLP) specifically, has progressed so much, that the great challenge of our time has become to try and put reins on the awesome capability these models provide – to ultimately derive value from their understanding of symbolism.
Looking to use AI and NLP? We have an army of practitioners with over two-years experience implementing NLP models, like GPT3 in production-grade systems.
Contact us to learn more.