Superintelligent Machines

ChatGPT4 is estimated to have an IQ of around 155. Albert Einstein had an IQ of about 160 and the average human IQ is around 100.

Schools may shift focus from rote learning to imagination, motivation, and kindness.

AI models already outperforming humans on some benchmarks:

• Standardized reading tests
• IQ pattern recognition tests
• Estimated verbal IQ scores in the top 0.1% of population
• Theory of mind and social reasoning at 9-year-old level

Consultants using ChatGPT completed more tasks faster and with higher quality.

Over-reliance on AI risks mistakes and hampers skills development.

Artificial intelligence (AI) capabilities have rapidly advanced in recent years, leading some experts to declare that machines can now match or even surpass human intelligence.

Intelligence has long been prized as one of humanity's highest traits. But innovations in AI and brain-computer interfaces could soon make extraordinary intelligence commonplace.

A paper titled "The New Irrelevance of Intelligence " argues that neural lace and AI systems will make traditional intelligence obsolete. Neural lace is an experimental brain implant that could let people directly access vast artificial intelligence networks, granting instant expert-level knowledge on demand.

This direct brain integration could propel average IQs to unprecedented levels, dwarfing even today's rare genius intellects. Equally disruptive is the rapid improvement of AI itself. Systems like GPT-3 already display strong language and reasoning abilities, which are only improving.

This could sound dystopian, but it also presents exciting opportunities. With AI assistance, anyone could gain skills it takes geniuses decades to develop - ethics, strategy, creativity, languages, and more could be available instantly. Schools would shift from rote learning to imagination, kindness, and motivation.

According to Dr. Alan D. Thompson , an AI consultant who previously chaired Mensa International's Gifted Families program, AI models like ChatGPT outperform gifted individuals on certain metrics.

Some highlights demonstrating AI's quickening pace:

• AI programs scored higher than most humans on standardized reading comprehension tests designed for high school students.
• GPT-3.5 matched or exceeded humans on Raven's Progressive Matrices, a classic IQ test involving pattern recognition.
• ChatGPT achieved estimated IQ scores of 147-155 on verbal intelligence tests, placing it in the top 0.1% of human test takers.
• AI models solved complex theory of mind and social reasoning tasks on par with 9-year-old children.

A study from Harvard Business School and Boston Consulting Group found that consultants using ChatGPT significantly outperformed their peers on work tasks.

The study examined 750 BCG consultants across 18 typical work assignments. Those using ChatGPT completed 12% more tasks on average, 25% faster, with 40% higher quality output versus those not using the AI.

ChatGPT particularly helped lower performers - the bottom 50% saw the greatest productivity and quality boosts when using the tool. Those trained on prompt engineering also did better than unsupported users.

However, the study found risks in over-relying on AI. On tasks outside ChatGPT's capabilities, consultants made more mistakes despite being more productive. And delegating too many tasks to AI could hamper skills development.

1. IBM's Deep Blue defeated world chess champion Garry Kasparov in 1997.
2. Google's AlphaGo defeated world Go champion Lee Sedol in 2016.
3. IBM's Watson won against human champions on the quiz show Jeopardy! in 2011.AI models, such as Libratus and DeepStack , have shown success in playing poker against human professionals.
4. AlexNet excelled in image recognition tasks, participating in competitions like ImageNet Large Scale Visual Recognition Challenge.

ChatGPT has passed the following tests :

• Wharton MBA Exam: B or B- grade on the final exam in operations management.
• US Medical Licensing Exam: Passed all three parts within a comfortable range.
• College-level Microbiology Quiz: Nailed it, answering both basic and complex questions correctly.
• Law School Exams: Received a C+ on exams in four different courses at the University of Minnesota.
• Google coding interview for level three engineer
• Scored 1020 in the SAT exam

While AI still has limitations, its rapid gains in language, reasoning, and social intelligence suggest a machine "intelligence explosion" could be imminent. As AI capabilities approach and potentially surpass the full scope of human cognition, we may need to reevaluate how these emerging superintelligent systems are guided and controlled.

Connect with me on Twitter: