ChatGPT is a program where users can type in a question or a task, and the software will pull information from billions of examples of text from across the Internet, to come up with a response designed to mimic a human.
“One of the key features that sets it apart is its ability to understand and generate natural language. This means that it can provide responses that sound natural and conversational, making it a valuable tool for a wide range of applications.”
Or, so says the chatbot about itself — ChatGPT wrote the paragraph above.
It has been a long time since a software release has consumed the tech community as much as ChatGPT, the latest offering from OpenAI, the AI startup founded by Elon Musk.
This chatbot, trained on massive pools of data and now able to answer any query you might have, gained more than a million users in less than a week. Post after post on Twitter revealed the inanimate interface crafting eloquent, believable prose on whatever topic was asked of it. Economist Tyler Cowen even got it to write a passable poem in iambic pentameter about economist Thomas Schelling’s theory of deterrence for foreign policy.
ChatGPT is far from perfect. It struggles with facts from time to time, as Bloomberg journalist Joe Weisenthal discovered when he asked it to write his obituary. And The Atlantic columnist Ian Bogost rained on everyone’s parade by observing that the chatbot doesn’t “truly understand the complexity of human language,” ensuring that “any responses it generates are likely to be shallow and lacking in depth and insight.”
But to Bogost’s boss, Atlantic CEO Nicholas Thompson, those imperfections won’t hinder the disruption this technology poses to a key part of the internet: search.
That’s more than just a minor problem. Without references to source materials, how can you trust what an AI is telling you?
Think of it in Big Brother terms. What happens when AI succeeds at becoming the major source of information? Someone, somewhere is training the AI with information. What happens when those trainers aren’t fairly benign researchers, but people with an agenda to promote?
Or what happens when people with an agenda to promote flood the internet with their propaganda? Can that hack the AI and get it to start repeating their propaganda for them?
Keep in mind that the second scenario has already been proven to have some success with people. So it is almost certainly going to be tried on an AI.
All your questions are addressed in Mo Gawdat’s incredible book “Scary Smart.”
One of my top 3 reads for 2022 in non-fiction. Although, the future scenarios he discusses can go either way, he is adamant that we in 2022 understand that A/I has already left human intelligence in the dust, and that for better outcomes, we must pay attention - NOW - at what we are inputting.
Everything you ask here is in the book and it will confirm some of your worst suspicions while hanging some hope on the upper boughs of your decision tree.
I don’t even know where to begin. Well, as the ChatGPT “said” - “It is important for society to carefully consider these issues and develop a responsible approach to the use of AI technologies.” But seriously, how do we develop a responsible approach? How do you set limits to something still in development without any chance of knowing in which direction its further development will go? Things can get out of hand so easily that I don’t even dare to think about it. We had a chance through history to see how what originally was a good idea takes a bad turn in horrifying ways. Anyway, thanks for sharing, it is a very interesting topic and raises many questions. Questions we cannot answer, but cannot afford to ignore.
When it comes to replacing human work, I will have to agree with the following - “Technology tends to create jobs in unexpected areas as it takes jobs away”. There are numerous examples to support this statement, and as the time goes by, I can only imagine there will be even more.
That was funny and witty, I have to say, thanks for lightening the mood. Yes, I never thought I would be discussing AI in any other way than chatting about old sci-fi movies (since I avoid the new ones), yet here we are.
Can’t stop laughing. Thank you!