David A. Windham thumbnail

Artificial Intelligence

I joined the OpenAI DALL-E 2 API1 waitlist just like tens of thousands of others did last week.  It’s a natural language processing system to create images from text or a text-to-image synthesis neural network.  I signed up because I just want to play around with it a bit. I’d imagine there will be hundreds of public facing generators in the coming months making everyone an ‘AI artist’.  As much as I’ve bemoaned artificial intelligence2, I had a conversation yesterday with some non-technical folks that seemed pretty enthusiastic about it and it got me to thinking about how I might explain my take on it to anyone else.  

The fella who sent me the link to signup is a retired IBM systems architect whom I’m always arguing with about artificial intelligence. He’s pretty excited about natural language processing as it relates to aging and regularly builds models for his ‘personal assistant’… whereas I’m not particularly fond of depending on any technology.  My car, Alexa, and Siri might have me labeled as incorrigible due to my belittling statements as to their capabilities. I almost find their lack of capabilities and the promotion of their usage mostly entertaining. I’m not saying that voice commands and sorting large datasets aren’t useful, but I’m neither amused or encouraged by their usage as ‘artificial intelligence’. It feels mostly like Jared from Silicon Valley3.

Part of my attitude is because I’ve always thought that intelligence encompasses so much more than just knowledge, problem solving, and logic. I think our mental capability, perception, and understanding involve much more. Machines are artificial and we shouldn’t devalue the word intelligence by attaching it to them.  It’s just another sales pitch… IBM’s Watson4 cheated on Jeopardy in 2011 and IBM’s Deep Blue5 cheated in 1996 at chess.  Yes a machine can respond quicker to a buzzer and they used human instruction to intervene in the Kasparov chess match. With that said, almost any cheap computer can easily whip up on me in chess making it’s moves in fractions of a second. The term Artificial intelligence is only used for the stuff that seems incomprehensible or problems yet solved. There’s actually a term for just that entitled ‘AI effect’6.  Wasn’t The Modern Prometheus ( Frankenstein ) a story about artificial intelligence.  This is the best take on it7:

I suppose much of the existing technology becoming more pervasive in our lives could be considered artificial intelligence… my car parking itself, telling me what music I might enjoy, or my email sorting itself into threads. Artificial intelligence is not magic. It’s just computer programming, logic, data, and math. The real danger is not being able to understand it.  Even though my text generated art might say otherwise, I’m not entirely dystopian about it because it certainly can be helpful when implemented well.  And because most of my work revolves around the machines, I’m of the ‘work smarter not harder’ school of thought. If I can figure out a shortcut tool, I’m gonna use it as long as it doesn’t lodge itself between me and the process.  

When I started using Github’s CoPilot8 and I knew there was going to be a way to make use of all of that publicly contributed code.  When Github was acquired by Microsoft for 7.5 billion in 2018, I knew that Microsoft’s Visual Studio was going to be the first key to capitalizing on it. I was chatting with a colleague recently who was complaining about ‘how the kids these days are writing code’.  I informed him of another acquaintance who teaches AP high school math and computer science and what he is using to teach his kids which involves an intermediary software which makes it easier for them to understand the fundamentals. I think that in the near future, the fundamentals of programming will be low code as the software will eventually just evolve to parsing natural language to describe setting strings, variable and arrays. Even though CoPilot is impressive in it’s execution, the first pitch says it all… ‘Autocomplete as a feature’ in code editors is nothing new and I’ve been using various editor plugins to save time long before CoPilot.  One effect I’ve noticed is that when I go and grab an autocomplete library package for whatever I’m working on, I have the tendency to learn less about it. 

So perhaps the real consequence of this so called artificial intelligence is the reason why IQ rates are dropping in many developed countries9,10.  I believe that it’s giving us a false sense of technological improvement especially to those that do not understand it.  I learned that lesson some years ago when I started substituting memory for machine11. Time will tell… I’m always intrigued by modern research that incorporates using technology and programming to gather and analyze extremely large datasets that would be impossible without the assistance of computers. I do think these types of AI usage will absolutely be used to benefit our lives in areas like science and medicine. I just don’t think we should offer up our own intelligence as a sacrifice. 

Update: 22/11/08

Github CoPilot now has a lawsuit on it’s hands12 which also makes ask the same question about the usage of the images in the DALL·E image generator.  Even though the output is modified, AI relying on data which has copyright is going to be problematic.  

Update: 22/11/30

Open AI released a new model called ChatGPT. I played around with it for a bit and had a long discussion about it while generating chats and surfing the Discord channel to see other dialogues. This is the first question I asked:

Although some of the results are really interesting, I’m fairly confident that aside from web applications and automated phone handling systems, the next application is going to be the drive through window at fast food restaurants. Here’s my Open AI DALL-E 2 image for “HAL from 2001 a Space Odyssey serving hamburgers”:

Update: 23/04/01 – I wrote a second essay on artificial intellegence @ https://davidawindham.com/artificial-intelligence-2/ .

  1. OpenAI DALL·E 2 – https://openai.com/dall-e-2/ 
  2. Artificial Intelligence – https://en.wikipedia.org/wiki/Artificial_intelligence 
  3. Silicon Valley – https://en.wikipedia.org/wiki/Silicon_Valley_(TV_series) 
  4. IBM Watson – https://en.wikipedia.org/wiki/Watson_(computer) 
  5. IBM Deep Blue – https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer) 
  6. AI effect – https://en.wikipedia.org/wiki/AI_effect
  7. Young Frankenstein – https://en.wikipedia.org/wiki/Young_Frankenstein 
  8. Github CoPilot – https://copilot.github.com
  9. NBC News IQ rates are dropping in many developed countries and that doesn’t bode well for humanity – https://www.nbcnews.com/think/opinion/iq-rates-are-dropping-many-developed-countries-doesn-t-bode-ncna1008576 
  10. Stanford MAHB – Idiocracy: is the decline in human intelligence undermining democracy? – https://mahb.stanford.edu/blog/idiocracy-is-the-decline-in-human-intelligence-undermining-democracy/
  11. Never Substitute Memory for Machine – https://twitter.com/windhamdavid/status/742432751987609600
  12. The Lawsuit that Could Rewrite the Rules of AI Copyright – https://www.theverge.com/2022/11/8/23446821/microsoft-openai-github-copilot-class-action-lawsuit-ai-copyright-violation-training-data
  13. OpenAI ChatGPT: Optimizing Language Models for Dialogue – https://openai.com/blog/chatgpt/