AI: don't be fooled by the "hype" that I got

Before you spin a lazy post or useless image, I think it is time...

that at regular intervals, we pause and reflect on how to be more effective.

Then tune and adjust accordingly.

(rings a bell?! - Agile Manifesto)

Time. Is. Now

Confronting the AI hype costs

• Data centre power surge six-fold in 10 years Capacity Concerns https://buff.ly/3IUxID2
• #AI & quantum computing driving spike in energy use
• Electricity used by data centres has risen by 400%
• the Internet alone is responsible of 3.7% of the global co2 emissions equivalent to the amount produced by the entire airline industry these figures are expected to double by 2025 https://theshiftproject.org/wp-content/uploads/2019/03/Lean-ICT-Report_The-Shift-Project_2019.pdf
• Current-generation AI models like ChatGPT now use over 500,000 kWh per day to operate—compared to 29 kWh for the average U.S. household. https://facilityexecutive.com/demand-for-data-centers-ai-puts-pressure-on-u-s-power-grid/
• Training alone for the GPT-3 model once at large scales uses 1,287 MWh, which is sufficient to power an average U.S. household for 120 years.
• The grid capacity can't handle the forecasts.
• While a single Google search requires half a milliliter of water in energy, ChatGPT consumes 500 milliliters of water for every five to 50 prompts.https://lnkd.in/eDs9PZeT
• On average, data centers are estimated to use between one million and five million gallons of water per day, equivalent to the daily demand of 10,000 to 50,000 residents.

We know that the accumulation of data alone does not drive better decisions.

We know that increasing noise or output churn does not get us to destination.

So how can WE better respond to change?

Costs do not stop there, AI is responsible for the severity of consumption of natural resources (including rare Earth) and human "labour"/slavery/trafficking

• Water costs https://lnkd.in/enpkjdQv
• The Hidden Carbon Cost of AI: Addressing Emissions in Machine Learning https://www.dhirubhai.net/pulse/hidden-carbon-cost-ai-addressing-emissions-machine-planet-davaajav-mix5c/
• Decisioning & Future of work: https://www.dhirubhai.net/pulse/large-language-models-shaping-future-work-product-ethics-ines-garcia/
• Consumption: https://www.dhirubhai.net/posts/gerry-mcgovern-07876469_the-ai-industry-alone-could-consume-as-much-activity-7178667732576436224-gfFR

(Hypocrite warning for Sustainability posts: https://lnkd.in/egrFZCQk)

Before you switch a feature on, ask yourself:

• What are the implications?
• What are the decisions you can consciously take?
• What are the better options?

There is also better ways to architect large data models (yes including LLMs) and also how you use it, and how much (and IF) is to be applied.

I'd tend to agree that the boom of usage over this year has been mostly by tapping into existing models, and unfortunately about only about a handful of options.

There are also some industry-specific or task-specific which are much more lean on processing.

Even when you tap on existing models, there are decisions such as reusability, data feed, storage etc that can be accounted and designed out for.

This is a big subject and much to unfold. But for now my message to us all is use your own processing power (brain) over lazy gpt posts or meeting summaries which are not accurate and no one reads.

Get into ACTION:

• Listen & check out > ccf azure carbon footprint thank you Green Software Foundation (environment variables). https://lnkd.in/er68zqK9
• Measure measure measure even leveraging AI?>
• CO2 AI www.co2ai.com
• CodeCarbon https://codecarbon.io
• Inmerse and inform yourself with Sasha Luccioni and the big science initiative bloom open llm https://lnkd.in/eGbv9Y7A
• Advance and democratize artificial intelligence through?open?source and?open science.

• As a champion for the Green Software Foundation I highly recommend yu to head over their work as well to the Impact Framework, readily available for you and your teams to tap (and expand) into: https://if.greensoftware.foundation/ their SCI Specification achieved ISO Standard!
• Also, get the FREE course and cert of Green Software, hosted at the The Linux Foundation https://training.linuxfoundation.org/training/green-software-for-practitioners-lfc131/

Most importantly, YOU are sitting at the table,

YOU are making many decisions a day,

YOU are greatly placed to be part of the solution.

Let's not conform to the conveyor belt of life, going through the motions without emotions. From?GPT?to?GTP: Genuine Truthful Purpose.

The future is not to be predicted nor forecasted, but to be imagined so that can be created.

It's up to us.