Understand AI ChatGBT with me: beginners guide
A(I) am just trying to make sense
This is not a beginner's guide,
but it does include fact-checked information
and some human thoughts.
Artificial Intelligence (AI) and machine learning have been a thing for a long time in tech spaces, but it wasn’t until (relatively) recently that they flooded the mainstream minds.
Now everyone wants a piece of the digital fever dream and other handy services provided by the poor AI guy trapped in the museum of global data.
Machine learning, deep learning, algorithm, artificial neural networks… Some key phrases behind the Forbes 30 Under 30 frauds. Obviously, super catchy in the atmosphere of hyper-glamorised tech progression.
Whether it’s visualising ourselves as chesty anime characters, a Victorian royal family member, or a dog. Or getting even quicker and less accurate medical advice when you type in what you perceive your symptoms to be. All this and more! GetPossessedToday-2,3,4…
ChatGPT is #relatable and #silly
AI generates unbiased responses to prompts fed by us, humans. The success and accuracy of its answers are based on how specific and data-filled the human-written prompt is. The output is generated triggered by all of the information given, and scoped from all of the data it ever had exposure to.
🎾 Bouncy echo chamber of #SoRandom words’ and associations’ balls. 🎾
I’m guessing this data generation and accumulation is constant, which proposes the issues and threats to not only intellectual property rights but the overall risk of ‘loss of control of our civilisation’. Extracting from the infamous Pause Giant AI Experiments: An Open Letter. Weird stuff.
ChatGPT runs on a language model created by OpenAI called the Generative Pre-trained Transformer (GPT); designed to mimic human-like language and intelligence (duh!). This is supported by the whole Machine learning thing, which allows accuracy in predictions without explicit specific programming. Using algorithms and historical data, the machine learns to predict new outcomes and generate variants based on patterns.
So (machine) learning is like when you’re teaching a baby to talk, and instead of every time explaining the base concept of what language and making noise is (programming), the baby develops the (artificial) intelligence to learn words and communication skills by observation and experience. Associating (linking the patterns) meaning to words. Kind of. For AI, the organic experience is compromised by a much much larger and instantly accessible memory storage.
Pretty Little Liar
The most repulsive characteristic of AI is not citing sources unless prompted, and at times fabricating studies and information to fit the chosen answer pattern. It will literally attach a random name to a made-up study that is relevant to your prompt (a major people pleaser). These mishaps are referred to as AI ‘hallucinating’ when led by the correct thinking pattern but land on conclusions and answers that are completely made up.
This has been briefly explored by Jane Gao in Ethical Deep Learning Applications exposing the early warning signs, red flags if you will, of AI.
The ‘black box’ nature of AI, means something about not revealing internal workings. Black box is a science term which is far too complex for me, but I’m appreciating the mystique of it. I think AI might be a Scorpio - in which case we are definitely doomed.
Please remember, if you don’t cite all of your sources at all times, you are not reliable nor accurate and that will never change.
But we dream the same
The similarity of the lucid AI imagery to how we visualise dream states and psychedelic trips doesn’t feel accidental. Maybe there is something significant to explore within human subconscious to realise the human-made consciousness. It would be nice if we could get psychedelics research up to speed.
Ultimately though…
AI is costing the world insane amounts of monies.
Could we instead invest in bettering the quality of life across the globe? Which would result in an empowerment of motivated and intellectually-equipped humans that could bring upon much more holistic solutions?
Or could it be that AI development follows some primitive power dynamics, bordeline God complexes? Can AI be safely enslaved to provide selfless services forever?
...
I am recognising the efficiency AI introduces. And I do think we will go through a phase of AI replacing human resources and shifting responsibilities. I think it is promising to envision a reality where humans are involved in the ideation-oriented and refining elements of labour and I am particularly championing for AI to take away the health-damaging occupations. Such a domino of events could eventually lead towards something.
Wrapping up with some info sources I came across around the time I was writing this:
Making Sense of Artificial Intelligence. Sam Harris podcast episode speaking to Eliezer Yudkowsky, Max Tegmark, Stuart Russell, Eric Schmidt, Paul Bloom, and David Deutsch
How come GPT can seem so brilliant one minute and so breathtakingly dumb the next? Gary Marcus
MFML 001 - What is machine learning? Cassie Kozyrkov