I like to test the limits of software so I asked ChatGPT what are its limits. Here is the answer.
I’m a large language model, which means my abilities and knowledge come from patterns in data I was trained on—up to a certain cutoff date. Consequently, I have several limitations:
No Real-Time Updates
I don’t have continuous access to the internet and can’t pull up new information in real time.
My knowledge is based on what was available at the time my training data was gathered.
Potential for Gaps or Errors
While I do my best to provide accurate answers, I may produce misinformation or omit important details if the information wasn’t well-represented in my training data.
It’s always a good idea to double-check critical facts with up-to-date, reputable sources.
No Personal Opinions or Experience
I don’t have personal experiences, emotions, or beliefs. My responses are generated from patterns and information in my training data, not from personal perspectives.
Can’t Provide Official Advice
I’m not a lawyer, doctor, financial advisor, or therapist. I can offer general information, but for specific, serious issues, it’s important to consult qualified professionals.
Potential for Bias
Like all AI models, my outputs can reflect unintended biases present in my training data.
I attempt to be neutral and fair, but I’m still limited by the text I was trained on.
Compliance and Safety Restrictions
I’m programmed to follow certain policies. For example, I can’t provide disallowed or harmful content, and I must avoid facilitating dangerous, unlawful, or exploitative activities.
If a request violates these policies, I may decline to fulfill it.
If there’s ever something I can’t do or a request I can’t address, I’ll let you know. Otherwise, I’ll do my best to help within these constraints.There are increasing times where I have asked ChatGPT to do something and it said it couldn’t. Today for example when I asked it another question it said I had 25 responses remaining until it reset on the 9th. This isn’t a problem, but it is a silly limitation. I am paying $25 a month, why are they limiting the responses?
Also when I have asked it specific information it will give me the error “I can’t access the internet right now” even though search engine functionality is a promise of the tier I am paying. I have also seen the prompt “I can’t answer that right now” which is probably a capacity thing. All in all, too many errors for a simple text query service that you pay a premium for.
Now I know that OpenAI would say that they are losing money and they are not using the responses to train their model they claim. Even if I believe that, it is hard to believe that you are losing money when most of your functionality is missing or not working. If you aren’t in IT or use it daily, it makes it a hard sell to continue with ChatGPT paid version when free versions can work. Maybe not as well, but probably close enough for the average person.
We know in IT that the best product doesn’t always win. Often it is the free product that wins until the company goes bankrupt and another free product is available. I wonder if AI is now at a race towards the bottom with this level of service.