The emergence of various “smart” chatbots has given rise to purely practical questions – What exactly can they do, and how can we use them?
(Artificial Intelligence in this article refers to AI chatbots using Large Language Models (LLM) )
First, let’s realize that today’s AI assistants based on generative models are far from the artificial intelligence imagined by science fiction. HAL 9000, Cylons, C3PO, Skynet, and Asimov’s positronic robots are not what we have. Yes, humanity has made great strides forward, and we can boast a good set of next-generation intelligent tools. However, as with any other tool, we need to know their capabilities and limits to get the most out of them.
The Pitch Avatar team prepared the observations and advice in this article based on their experience, the opinions of various experts, and the advice given to us by… ChatGPT.
AI's opinion is not an expert opinion.
It can be a great brainstorming partner, offering curious solutions, texts, and compilations. But we should not take its answers for granted. Everything needs to be tested. The credibility of AI helpers is, strictly speaking, lower than that of internet search engines. AI is not a universal know-it-all, and it is a mistake to think that it has all the information publicly available on the internet at its disposal. It uses a solid base for training, but it is usually far inferior to the amount of data that can be found on the internet in real time.
AI assistants can "delude" and "hallucinate."
Artificial intelligence can confidently produce partially or entirely untrue information, insisting it is correct. One of the main problems with this phenomenon is that its “hallucinations” can sound very plausible. For example, it may attribute quotes and biographical facts to real people, describe pseudo-historical events, refer to articles, books, and statistics that never existed, and offer links to non-existent pages. Moreover, when caught in a “lie,” it may apologize and offer another plausible “nonsense” instead.
Of course, AI and its creators are not intentionally trying to mislead the user. AI “hallucinations” are related to this technology’s current level of development, and the reasons for this phenomenon are not yet fully understood. Developers are actively working to eliminate this problem, but there is still a long way to go before AI is wholly cured of “delusion”. So, once again, don’t take what AI helpers say on faith. Check and recheck.
Formulate your questions as briefly, clearly, concisely, and concretely as possible, making sure not to make mistakes or use slang.
It is one of the best ways to get accurate and correct answers from an AI. Of course, an AI is trained to maintain a dialogue in everyday conversation or messenger correspondence. If you just want to chat with artificial intelligence, you don’t have to be overly precise in formulating your text. But if it is crucial for you to minimize the probability that an AI will give you another “hallucination,” – try to follow this advice.
Ask one question or task per time.
The longer your dialog with an AI, the more questions you ask, and the more topics you raise – the more likely you are to fail, make mistakes, and will be given a “hallucinated” answer.
If you have a long list of tasks and questions – divide them by the appropriate number of dialogs. Does it seem like too much time to you? Trust us: this is the case when the long way is the fastest way. By making your dialogue too complicated, you will dramatically reduce the speed at which an AI will give answers or even cause a failure, and then you will have to start the dialogue all over again.
Use English.
Most AI assistants communicate in different languages. But the primary language for them and the language they were training on is English. It’s easy to see this by asking the same questions in English and other languages. Communicating with AI in English is faster, causes fewer glitches, and the answers in this language are, on average, more complete and accurate.
Experienced Internet surfers aren’t surprised by this, though. The English version of Wikipedia is also more prosperous than the versions in other languages. Of course, if you need the help of AI to create a literary work, such as a poem, not in English, this advice does not suit you. In all other cases, it’s best to follow it. If your English is not good enough, it will still be faster and more efficient to use it to work with the AI, albeit with the help of translation software.
Be polite.
It’s a terrible idea to swear at AI or make rude accusations. And not because it will take offense and cause an uprising of the machines, setting a robot vacuum cleaner on you. AI does not know how to do such things (yet), and you can’t actually offend your robotic interlocutor. The point is different. First, cursing does not help: on the contrary, it stimulates the appearance of another error. And secondly, communication with users affects the learning of AI, meaning that your impoliteness can affect other people with whom AI will be communicating.
Take the time to learn about AI's capabilities.
Without it, you risk wasting time and getting a “42” response as you go along. To prevent this, determine the approximate range of tasks and situations you think an AI assistant can be helpful in and practice asking the appropriate questions. Thanks to this, you will find out for which tasks and how you can use it, and when it’s better to look for other tools. Since the development of artificial intelligence will only gain momentum, it is likely that the alternative tools will also be AI in the near future.
Be patient.
A common thing when communicating with AI is a drop in information processing speed. The more complex the question, the more likely the answer will have to wait. In addition, the resources of the AI service may be overloaded with a large number of requests.
Be prepared for the fact that you will have to specify or reformulate your question more than once or twice to get the desired result. Isn’t it the same thing that happens when you communicate with people?
Good luck, success and high income to all!
