AI's Explosion
Modified: 2024-03-25 10:31 PM CDST
text
AI History (Smith, G., & Funk, J. ((2024) When it comes to critical thinking, AI flunks the test. Chronicle of Higher Education.
- 1956
- Dartmouth College Summer Workshop: John McCarthy: the conference would “proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it.”
- 1965
- Herbert Simon:“machines will be capable, within 20 years, of doing any work a man can do.”
- 1970
- Marvin Minsky: in “three to eight years we will have a machine with the general intelligence of an average human being.”
- 2008
- Shane Legg: “human-level AI will be passed in the mid-2020s.”
- 2015
- Mark Zuckerberg: “one of [Facebook’s] goals for the next five to 10 years is to basically get better than human level at all of the primary human senses: vision, hearing, language, general cognition.”
- 2022:
- Bill Gates: ChatGPT and other large language models (LLMs) are “every bit as important as the PC, as the internet.”
- Jensen Huang: ChatGPT is “genuinely is one of the greatest things that has ever been done for computing.”
- Geoffrey Hinton: on ChatGPT “I think it’s comparable in scale with the Industrial Revolution or electricity — or maybe the wheel.”
- 2023:
- Chris Anderson: “Surely it’s not crazy to think that sometime next year, a fledgling Gemini 2.0 could attend a board meeting, read the briefing docs, look at the slides, listen to everyone’s words, and make intelligent contributions to the issues debated?”
Critical Thinking
- Robert H. Ennis defines critical thinking thusly:
- “reasonable, reflective thinking that is focused on deciding what to believe or do.”
- He gives the following critical thinking skills:
- Open mindedness
- Being well-informed
- Assessing the credibility of sources
- Identifying conclusions, reasons, and assumptions
- Judging arguments
- Defending reasonable positions
- Asking the right questions
- Knowing how to hypothesize and experiment
- Defining context specific terms
- Cautiously drawing conclusions
- Using all of the items above correctly
- Smith and Funk claim that LLMs cannot perform any of the tasks above
- They had three LLMs: ChatGPT (OpenAI), Copilot (Microsoft), and Gemini (Google) answer prompts for the following areas:
- The answers given were wordy and incorrect
- The answers showed no evidence of critical thinking
- They concluded:
- "we are still a long way from the goal defined by AI boosters like McCarthy, Simon, and Minsky, not to mention the breathless hype emanating from Silicon Valley. For the time being, it looks like we’re going to have to continue to do our critical thinking for ourselves, and teach our students to do the same."
- Tell class how you have used a LLM such as ChatGPT
Consciousness
- We all experience it, but no one knows what it is
- So, how can we imagine machine consciousness
- Many of the disasters of science fiction result because of bad programming.
- RoboCop
- ED-209 malfunctions during test, killing the subject who had dropped his weapon
- SkyNet
- System gained consciousness and initiated nuclear attack
- HAL 9000
(video)
- Open the pod bay doors, Hal
- I'm sorry, Dave, I can't do that.
- DishBrain (video) Real Science
- DishBrain is about 800,000 neurons grown in a rectangular petri dish
- It learned to play Pong by itself after being given motor "cortex" and a sensory "cortex"
- NO RULES WERE PROVIDED
- In five minutes it was playing Pong at a better than chance level
- Is this consciousness?
- Is it a machine?
AI Takeover
- What about AI Takeover?
- Old science fiction theme
- Frankenstein (1818)
- He wants a wife (video, 1935)
- Superintelligence
- Bostrom: "any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest"
- Unfriendly vs Friendly
- Alignment
- Make AI values and human values align
- Part of AI Safety
- Reward Hacking
- AI system achieves programmed goal but does so in a way not intended by the programmers
- Fitness functions are programmed in
- Humans, animals, and AI systems may "discover" new and unforeseen methods of achieving the fitness function
- Chimpanzees are difficult to toilet train
- NASA wanted to train chimps to use toilets for prolonged space flight
- When on the toilet, chimps would receive food pellets when feces dropped into the toilet
- One chimp learned to save one pellet and then drop it down the toilet
- Reward Hack: the chimp now received more pellets, not what the trainers had designed as the fitness function
- A. J. Riopelle, personal communication
- Do you worry about AI Takeover?
Military AI
- Autonomous Weapons (video) scroll down to video
- AI Weapons (video)
- Most current AI weapons have some degree of human control
- But, how long will that last?
- Drone Pilots
Swarming Drones
- Swarming Drones have many purposes
- Military
- Overwhelm anti-drone defenses
- Kamikaze drones
- Agriculture
- Emergency management
- Entertainment
- Control
- Centralized
- Decentralized
- Autonomous
Geometry and AI
- Alpha Geometry (video) (Google Deep Mind)
- Geometry, while the oldest part of mathematics has proven to be the least amenable to AI solutions
- Uses
- Neurolanguage model and Symbolic Deduction Engine
- Solved 25/30 problems in the IMO competion
- Mean was 15..2
- Gold human winner solved 26
- Silver human winner solved 22..9
- Bronze human winner solved 19.3
These are some of the newer ways AI has evolved
Back to Main Page