Why AI Bots are Batsh*t Crazy
Though this was awhile ago, what goes on the internet stays on the internet, forever. And plus, how can you not gawk at these tweets? Especially from an artificially intelligent bot who's objective is to learn to talk like humans.
Tay was a Microsoft AI Tweet bot as an experiment in "conversational understanding." The more you chat with Tay, said Microsoft, the smarter it gets, learning to engage people through "casual and playful conversation."
It took 24 hours for humans to corrupt this bot. Leading to the tweets above.
Should Machines Really Learn from Humans?
Clearly, the benefits of a potential artificial intelligence is highly intriguing to humans. Being able to accumulate knowledge and information from the masses and combining it into a collective higher intelligence.
Like the Guardian short video points out, this isn't without moral and ethical concerns. In fact, the biggest problem here, is one of values. Until all of humanity agrees on the same core values, that we can then program to be the ultimate purpose to an AI. Bots and artificially intelligent machines will face the same existential problems that humanity has faced since the beginning of our consciousness, what is their purpose?
The simple way to put it, is that our society has become incredibly complex. We have different sets of written and unwritten rules that people follow, some by choice, some by necessity. For example, humans need to get sleep every night in order to function at an optimal capacity, but this limitation doesn't exist to machines. We have created institutions such as law, economics, government and schools as a way to structure our collective existence. Meaning if we don't provide a set of hard guidelines and restrictions for AI to abide by, they will ultimately create their own set of rules that makes sense to their limitations.
Magnifying Human Biases
An important factor to consider in the progression of AI is that they need to learn from someone. As machine learning takes all of it's original information and data points from humans, this will only further perpetuate existing human biases and flaws that are deeply embedded in us. This infographic shows all the documented cognitive biases that we have developed due to our limited neurological capabilities.
This means that in order for an AI to actually be representative of humanity as a whole, the amount of initial information we feed it needs to be an even distribution and representation of people from all societies and backgrounds. If we only provide a machine to learn from a group of people with conservative values, then it would continuously optimize towards only those values. And given the diversity of societies as a whole, you can easily understand why this may become an issue if not set up properly since the beginning. Which is also why highly influential figures in the tech space such as Elon Musk, deems that true artificial intelligence needs to be democratic, leading to projects such as OpenAI.
How Bots can Support Humans
Don't worry, AI bots at their current state have yet to pass the Turing test, meaning that people can still determine them to be machines, not real humans. Many bots exist for commercial reasons, such as talking with customers online as traditional customer service representatives did, automating the repetitive and mundane tasks that are now programmed.
The more interesting bots on the other hand, are the ones that are making a personal impact on humans. Being used in industries such as psychology and mental health.
Replika is a bot that users talk to, who listens and learns to talk like you, essentially becoming a replica of you in chat form. This app showcases an unique situation, and a critical juncture in the bot evolution. The story of this company's origins is fascinating, as it's a story about overcoming mortality. More interestingly, is the fact that individuals now feel like they can trust and open up to a machine more than people. Meaning machines can now replace humans in aspects more than just work and labour, but also human connection.
There is a ton of progress being made everyday in the world of algorithms, machine learning and artificial intelligence. However, the same roadblocks to developing true AI looks very similarly like the roadblocks of humans and society. So to answer the question, why AI bots are batshit crazy, it's because we also have yet to figure out the answer to our existence. The field of machine learning has a lot of the same cross-overs to the humanities, and the development of AI will push and reflect the development of society and people.
If you found this information helpful, consider giving out email list a try. We only send 2-4 emails a month when new content comes out.