We are currently on the verge of yet another societal “revolution” that will exert an unprecedented impact on our lives. It may surpass prior seismic cultural breakthroughs like the internet, smartphones, and social media. Artificial intelligence (AI) has been fermenting for several decades, gathering steam to become equivalent (and eventually superior) to human intelligence. The escalation of AI sophistication will be jarring and perhaps change human life in completely unpredictable ways.
Composing thoughts into words and coherent sentences has always been a uniquely human attribute among all living organisms. Now, that sublime feature of the human mind is being simulated, thanks to advances in AI software, ironically created by the human mind itself! On November 30, 2022, Open AI introduced ChatGPT (generative pre-trained transformer), which can generate an article on any topic a user requests. Within a few weeks, it was used by more than 100 million people. ChatGPT is taking the world by storm because it is a harbinger (some pessimists may label it an omen) of how human existence will be radically impacted in the future. Such AI breakthroughs to surpass human intelligence are, ironically, the product of the advanced human brain, which I previously described as concurrently a triumph and a blunder by evolution.1
How we got here, and what’s next
ChatGPT is a large language model based on neural networks.2 It generates realistic text responses to a wide range of questions by mimicking the pattern of language in gargantuan online databases. One Hong Kong–based, AI-powered drug discovery company (Insilico Medicine) declared it published articles generated by AI tools, even before ChatGPT became available. This indicates how AI can be misused in scientific publications and may be hard to detect as a new form of plagiarism.3
The roots of AI date back to the 1950s, when Alan Turing, now considered the father of AI, published a seminal article about creating a machine to “imitate the brain” and to “mimic the behavior of the human.”4 The term “artificial intelligence” was coined in 1989 by McCarthy,5 who defined it as “the science of engineering for making intelligent machines.” Since then, several subsets of AI have been developed:
- Machine learning: The study of computer algorithms to generate hypotheses
- Deep learning: A type of machine learning algorithm that uses multiple layers to progressively extract higher-level features from raw input. (Both machine learning and deep learning are used in the burgeoning fields of computational psychiatry6 and neuroscience research7)
- Expert knowledge system: A computer-based system that mimics human decision-making ability
- Neural networks: An interconnected group of artificial neurons that uses a math or computer model for information processing
- Predictive analytics: An algorithm to predict future outcomes based on historical data.
These subsets of AI have been used to identify psychiatric disorders using neuroimaging data8 and to classify brain disorders.9 There are many potential uses of AI in psychiatry.10,11 My first experience with AI was 13 years ago, when we conducted a project to distinguish fake suicide notes from genuine ones.12 AI was more successful in correctly identifying fake notes (78% correctly detected) than senior psychiatric residents (49%) or even faculty (53%).
AI will dramatically change how humans interact with the world and may lead to enhanced creativity and new explorations and forays into novel, previously unknown horizons. It is expected to significantly boost the global economy by many trillions of dollars over the next decade. Major high-tech companies are vigorously competing to develop their own AI tools like ChatGPT (Microsoft invested $10 billion in Open AI). Google, which owns DeepMind (an AI lab that invented the T in GPT) developed its own chatbot called Bard. Amazon has invested heavily in Stability AI by giving its founder and CEO Emad Mostaque 4,000 Nvidia AI chips to assemble the world’s largest supercomputer (1 year ago, Stability had only 32 AI chips!). Apple recently integrated Stable Diffusion into its latest operating system. Chinese tech giants Alibaba and Baidu also announced their own chatbots to be released soon.
Other competitors include Cohere, Hugging Face, Midjourney, GitHub Copilot, Game Changer, Jasper, and Anthropic, which released Claude as its chatbot at a lower cost than ChatGPT. Open AI also developed Dall-E2 in April 2022, which can generate very realistic images from text, one of which recently won an award at an art competition.
Continue to: One of the major...