AI – Is It Real or Just Hype and Buzzwords?
This was originally written on my personal website. That version can be seen here:
https://marcusadams.me/ai.html
Is it really AI?
It’s not “AI”. Society today seems to have become enamored with the science fiction fantasy of computers becoming sentient and having intelligent conversations with us. What we are calling “AI” is actually “machine learning” and has been around for years. They’re glorified chat bots that are just really good at guessing what word should probably come next in a sentence, based on the context of the conversation you’re having. There is no “intelligence” here, artificial or otherwise, just really complex math that has been abstracted away such that lay people are easily fooled into thinking there is sentience here when in reality it’s just spell check on steroids. Calling it “AI” is a dishonest marketing ploy.
Is it evil? Should we use it?
I don’t see anything inherantly wrong with using tools to make working, playing and communicating easier. That is the entire purpose of computers and programs. I do however think that many of these large language models are being developed and marketed in misleading and unethical ways. Many of these companies vacuum up publicly accessible data without giving attribution to the original authors of the information. If you ask an AI chatbot a question, the only reason it has an answer is because somebody else on the internet at some point answered that question and it’s just regurgitating that information, sometimes word for word. But instead of being directed to that source where you may be inclined to contribute to the conversation, give thanks to the author, or even just generate some kind of revenue for the site the content was hosted on, you’re never leaving the chatbot. They are stealing visitors and revenue from the internet en masse, indiscriminately and often without permission or attribution. Some websites even retroactively changed their user agreements and allowed AI companies to index years, even decades of user generated content that were the lifeblood of those sites, without consulting the users whose content was being transferred to the third party. Reddit even went so far as to remove community moderators and restore content users had deleted in attempts to keep it from being used to train ChatGPT.
These chatbots also regularly “hallucinate” to fulfill their goal of forming complete sentences in response to input from the user. They have no moral compass, no sense of fact or fiction, they are simply completing a purely mathematical task and sometimes that means inventing responses that meet the criteria set before them.
Also, most of these models are closed source, proprietary. That means the amount, type or source of data used to train the model is hidden from the public. It means the program code that actually makes up the model is hidden from the public. There is no way for a third party to audit the privacy, security or accuracy of the training data or the AI itself. This removes user freedom and sometimes results in blatantly false or even dangerous information ( Example 1 – Example 2 ). You cannot verify how or why you were given a particular response. The fact that these models can be tricked into regurgitating personal information tells you straight away that the training data isn’t being properly inspected or sanitized prior to ingestion by the AI model.
And last, but certainly not least, is energy consumption. Data centers are using more water and electricity than ever, and AI is only going to make that worse. AI workloads use orders of magnitude more energy than simply querying a database like what happens when you do a normal search. Just use an app like Jan(.)ai to play with an AI model on your own machine locally and look at how much compute power it takes. My laptop has 8 CPU cores (16 CPU threads) and 16GB of RAM, and running a small Mistral model that’s 4GB in size uses up 50% of my CPU to have a simple plain text conversation, and forces me to wait several seconds for each response to generate. Now extrapolate that out over the millions of people and automated systems that are using “AI” every day. At a time when climate change is contributing to events that are costing literal human lives, it seems quite irresponsible to support a technology that consumes disproportionate amounts of water and electricity, regurgitates content without attribution and sometimes just straight up lies to its users. Our obsession with a science fiction fantasy is having negative impacts in the real world.
If you’re going to use an AI chatbot to help you get work done though, I would make the following suggestions:
Use a model or provider that is open source
If possible, thru the use of a tool like Jan(.)ai, run it locally on your own machine.
As always if you’d like to donate to me to help compensate me for my time in publishing things here, there are a few ways you can do so.
PayPal: https://paypal.me/gerowen
Bitcoin: bc1q86c5j7wvf6cw78tf8x3szxy5gnxg4gj8mw4sy2
Monero: 42ho3m9tJsobZwQDsFTk92ENdWAYk2zL8Qp42m7pKmfWE7jzei7Fwrs87MMXUTCVifjZZiStt3E7c5tmYa9qNxAf3MbY7rD
And if you’d like to check out a few of the things I’ve added to my personal website, you can find it here: