Consider This When Choosing AI

Disclaimer: I try to educate myself, but these are my opinions and decisions for usage based on my current knowledge. If other research and sources refute what I say, please feel free to comment (politely) with them. I’ve included some links for each section in the comments.

The topic of AI has been nagging at me for some time now.

Kind of literally.

?? ?First I attended Amy Frushour Kelly ’s recent webinar on the Ethics of AI.

?? ?Next came the announcement of my upcoming webinar with ACES: The Society for Editing about AI in editing tools.

?? ?Then this morning I saw Jonathan Jordan ’s post on AI and received Mignon Fogarty 's AI sidequest newsletter.

AI hasn’t gone away as a hot topic.

So I’d like to hopefully give you some things to consider when deciding whether to use AI or not (spoiler alert: this is strongly from the NOT perspective).

I really tried to make this a post, but I went over the word limit ??♀?

?? ?AI Uses Significant Amounts of Energy

In a world that’s trying desperately to reduce human impact on our planet, the energy usage and CO2 output of AI is simply staggering.

Consider this: Training ChatGPT-3 took the annual energy of 130 US homes. Training ChatGPT-4 used 50 TIMES that amount. And that’s just one company. The rapid advancement of AI computing means the energy required to run it is doubling every 100 days and is “projected to increase more than a million times over the next 5 years.” ??

Companies using AI are clearly stating that they don’t care for the environment, and the emissions resulting from their AI usage should be added to the total for their company and count toward their net zero totals.

On their own website, Canva states: “The scientific consensus is clear that we must limit global warming to 1.5°C. To achieve this, the planet needs to drastically reduce climate pollutants and achieve net-zero emissions by 2050.” Yet, Canva, just last month, bought Leonardo.ai, a popular generative AI startup. How they plan to reduce their emissions to net zero while increasing AI usage is beyond my understanding.

In short: every AI prompt costs our planet even if it lowers the cost on your wallet.

?

?? AI Usage Is an Ethical Nightmare

When it comes to using AI, there is also an ethical component. One opinion seems to be that using AI for ideation is fine but having it produce a final output, such as writing a whole blog to publish, is unethical. If the writer or blog poster declares it was written with AI, does that suddenly make it ethical?

Amazon is requiring people who publish through KDP to declare whether AI was used in the creation of the book, but as of yet, they don’t show this information to the consumer. Is that ethical?

And then there are the AI data sets. Can any usage of AI be considered ethical if the model it uses was trained on stolen data? There are 25 major course cases against significant AI players right now. And more companies are jumping on the bandwagon of selling their data without the consent of the users who put it on their site (I’m talking to you, Reddit).

?

?? AI Is Impacting Our Social Skills

Now let’s consider the societal implications of using AI. I’ve recently come across a trend where users of Instagram ask ChatGPT to examine their feed and then “roast” them with one paragraph. In the world of social media, where there’s already enough negativity, why ask AI to add more? I don’t understand the appeal.

AI affects our social skills in other ways too. LinkedIn recently released its feature offering AI assistance to write comments on posts. On a social platform driven by human connection, AI is offering a social “cheat” to interacting with others—or at least, appearing to interact. Not to mention LinkedIn trying to appear as if they interact with users through their “collaborative articles,” which are created by AI and often are nonsensical.

?

??? AI Spreads Bias and Disinformation

AI is increasing the amount of false information on the web (and will keep feeding on itself until it believes its own lies). Recently an AI-driven bot farm was exposed in Russia that was spreading pro-Russian content filled with disinformation. They say the farm was able to produce more information faster, which was also more convincing, than without AI usage. This is especially concerning from a global security perspective, and yet 72% of UK residents in a 2019 study weren’t even aware of what a “deepfake” was.

AI is also highly biased, with AI image generators being among the worst offenders. One analysis of 5,000 AI-generated images from the program Stable Diffusion found that “it takes racial and gender disparities to extremes—worse than those found in the real world.” AI also feeds from its inputs, meaning if harmful and stereotyped material is put in, it’s more likely to skew the resulting text. The advent of “personalized” generative programs will only perpetuate our biases or continue to feed us wrong information.

?

?

In conclusion, there is a lot to consider when it comes down to using generative AI. I personally use it very sparingly and only for idea generation, never output.