AI and the Energy Cost to Do Our Thinking

The human nervous system runs on about 20 watts.

According to this article, training GPT-3 required as much energy as powering 120 houses for a year (before adding the required energy for the chatbot to reply to users’ prompts), and GPT-5 must be far, far higher.

In fact, all across the country is a quiet initiative to build new power and processing centers that is needed to keep up with the demand. As adoption rates increase and AI capabilities advance this will become a mandatory component of overall AI infrastructure.

Everything comes with a cost.

I read that Chat GPT usage dropped over the summer, then spiked again in toward September. Why? Our children are using it for schoolwork.

I use it myself for many things including researching pricing and features of items I am interested in purchasing, and it saves huge amounts of time and hopefully prevents errors by ensuring compatibility. AI offers real value for certain aspects of life.

My fear is it is replacing our ability to think for ourselves in more critical ways. How long before people turn to AI for life advice instead of figuring out what is best for them? Let the AI tell us what we want; it knows all.

AI is a valuable asset and tool to be utilized for certain circumstances, but it needs to be carefully applied for others. Is it showing goods you are researching from the company that paid the AI provider over a better choice from another company? How would we ever know if it is?

As the cost for energy to run AI processing increases we must keep the human cost in mind as well.

What are your thoughts on AI and how it is being used, and how it might become used in the future? Please comment below.