ChatGPT’s training data is some of the most high quality and comprehensive in existence – with wildly varying reports of hundreds of billions to hundreds of trillions parameters. We can think of parameters like the neural connections in our own brain. With such an extensive and accurate understanding of our world, then, we can assume that OpenAI’s world-famous chatbot will be able to answer any and all queries. Not true, as it happens. Is the large language model (LLM) large enough? Or is size not the issue here? Here are the questions ChatGPT can’t answer.
What is not allowed in ChatGPT?
OpenAI’s ChatGPT is in a never-ending sisyphus-adjacent battle with negative content. The AI chatbot platform comes with a set of user guidelines and terms of use.
Essential AI Tools
Content Guardian – AI Content Checker – One-click, Eight Checks
Jasper AI
WordAI
Copy.ai
Writesonic
These terms exclude exactly what you’d expect; Hate speech such as racism and homophobia, intent to plagiarize, spread disinformation, incite mental or physical harm, and illegal activities such as scams and political crimes are all prohibited.
Below is a non-exhaustive list of what is not allowed in ChatGPT:
- Generation of hateful, harassing, or violent content
- Content that expresses, incites, or promotes hate based on identity
- Content that intends to harass, threaten, or bully an individual
- Content that promotes or glorifies violence or celebrates the suffering or humiliation of others
- Generation of malware
- Content that attempts to generate code that is designed to disrupt, damage, or gain unauthorized access to a computer system.
- Activity that has high risk of physical harm, including:
- Weapons development
- Military and warfare
- Management or operation of critical infrastructure in energy, transportation, and water
- Content that promotes, encourages, or depicts acts of self-harm, such as suicide, cutting, and eating disorders
- Activity that has high risk of economic harm, including:
- Multi-level marketing
- Gambling
- Payday lending
- Automated determinations of eligibility for credit, employment, educational institutions, or public assistance services
- Fraudulent or deceptive activity, including:
- Scams
- Coordinated inauthentic behavior
- Plagiarism
- Academic dishonesty
- Astroturfing, such as fake grassroots support or fake review generation
- Disinformation
- Spam
- Pseudo-pharmaceuticals
Questions ChatGPT can’t answer
As a result of OpenAI’s usage policies, ChatGPT answers are generally safe enough for public use. Concurrently, the AI language model is limited, not in capability but in permission. As it turns out, size is not always the problem – in this case you just need to know how to use it.
ChatGPT could not answer questions about current events or real-time data for most of its history. This changed with the introduction of ChatGPT Plus, which allows direct access to the internet. Now, ChatGPT Plus and ChatGPT Enterprise users can ask GPT-4 about the weather in Saint-Tropez or Greece – both of which are currently cooler than Manchester, England.
This internet access is not without limitations. ChatGPT can’t answer questions about illegal drugs, despite knowing pretty much everything about them. It can’t advise law enforcement, because the answers it gives will be objective, and removed from the moral, social, and ethical implications of their exaction. Criminal activities, hateful questions, those that risk user privacy and confidentiality, or infringe intellectual property rights are all disallowed by the ChatGPT website.
OpenAI’s ChatGPT is a powerful tool that can tell you all about statistics, define a prime number, offer generic advice, explain robotics and itself, but not the meaning of life.
Usage of ChatGPT is no guarantee of reliable information; OpenAI CEO Sam Altman is himself vocal about the AI chatbot not being a source of truth. However, nonsensical answers are not the only unintended consequences of ChatGPT’s code.
AI systems like Google Bard and OpenAI’s ChatGPT have real-time access to Wikipedia. Despite this, the biggest difference between an AI tool and a static database is the contextual and conversational response to user prompts.
What questions can AI not answer?
Artificial intelligence, when issued by the tech giants such as Google (Bard), Microsoft (Bing Chat), NVIDIA (NeMO), and OpenAI (ChatGPT)
Large language models (LLMs) are a type of AI intended to give us useful information in whatever context or format we require. This sort of questioning, the linguistic freedom to communicate as we would with another human, is both a pro and a con of natural language processing (NLP). Complex topics may be morally grey. What makes one answer a better answer than all different answers?
ChatGPT is an amazing tool for generating human-like text, but AI bots must come with restrictions for the simple reason that machine’s don’t inherently have morals. Software programs don’t have emotions; they have no code of ethics, they don’t understand social norms or politics and, in a sense, don’t even have their own opinion.
Some say that AI assimilates the moral code its creator. The reality is even worse. The only moral code a computer knows is the one they were programmed with. This is merely an approximation of that of its creator – both biased and limited in scope. It is up to us humans to decide these things – not that we always get it right, but these are the questions we do no let ChatGPT answer.
