Solutions
Pricing
Log in
Teams
Get started
Log in
Get started
Extra credit Macro
DH
by
Destiny Hadcock
·
11 views
·
Updated Mar 25, 2024
Share
Embed
Create your own
Audio Transcript
Auto-generated
00:00 - 00:06
Hello everyone. My extra credit assignment is over the environmental costs of A I.
00:09 - 00:12
What is A I or artificial intelligence?
00:13 - 00:16
A I is a technology that allows computers and
00:16 - 00:20
machines to stimulate human intelligence and problem solving skills.
00:20 - 00:22
A I can perform tasks that would
00:22 - 00:24
normally require human knowledge and intervention.
00:25 - 00:29
One of my personal favorite examples of A I is a robot named Aurora
00:29 - 00:33
located inside the sphere in Las Vegas, Nevada.
00:33 - 00:35
Although Aurora's creators spent quite a bit
00:35 - 00:38
of their time and dedication building her
00:38 - 00:40
A I doesn't have to be complicated.
00:40 - 00:46
For example, GPS, self-driving cars and digital assistance are all A I. Technology
00:50 - 00:52
A I creates concerns
00:52 - 00:57
A I is a power eater, meaning it takes a lot of resources to keep it running,
00:57 - 00:59
which has become a pressing issue.
00:59 - 01:03
Concerns about its environmental footprint have surfaced
01:07 - 01:09
energy consumption.
01:09 - 01:13
A is on a peaceable sensation for energy is a growing concern.
01:13 - 01:16
As A I's algorithms become more complex,
01:16 - 01:18
the energy demanded skyrockets,
01:18 - 01:22
the International Energy Agency or IE A
01:22 - 01:28
forecasts a doubling of electricity consumption by data centers and A I by 2026.
01:28 - 01:30
Highlighting the magnitude of this challenge,
01:31 - 01:32
tech giants such as Amazon,
01:32 - 01:37
Google and Microsoft have seen dramatic increase in energy usage.
01:37 - 01:40
Raising alarms about sustainability.
01:40 - 01:41
In particular,
01:41 - 01:44
the development of new A I tools like Google's
01:44 - 01:49
Bard and Microsoft's copilot contribute significantly to this search
01:52 - 01:53
water usage
01:53 - 01:55
alongside energy concerns.
01:55 - 01:59
A I's intensive water usage poses another environmental threat.
01:59 - 02:02
In Google's latest environmental report.
02:02 - 02:06
2022 the company consumed 5.6 billion gallons of water
02:07 - 02:08
to put it into perspective.
02:08 - 02:13
This equates to nearly 10 days of water for the entire city of London.
02:13 - 02:14
Additionally,
02:14 - 02:18
Microsoft consumed nearly 1.7 billion gallons of
02:18 - 02:20
water from its operations in 2022.
02:21 - 02:25
Research from the University of California riverside found that the source
02:25 - 02:29
to power Open A I Chat GP T made by Microsoft
02:30 - 02:38
consumed 700,000 L or 100 and 6400 gallons of clean fresh water.
02:39 - 02:43
Chat GP T needs to drink a 500 mL bottle of water
02:43 - 02:47
to hold a conversation that includes 20 to 50 questions and answers.
02:47 - 02:52
The company has a newly launched Chat GP T four. That is a larger model size.
02:52 - 02:55
All these numbers relating to water consumption are
02:55 - 02:58
expected to increase due to this newly launched platform
03:01 - 03:03
carbon footprint.
03:03 - 03:06
In 2019, University of Massachusetts, Amherst,
03:06 - 03:10
researchers trained several large language models and found
03:10 - 03:12
that a single A I model can emit over
03:12 - 03:18
626 £100,000 of CO2 which is equivalent to the
03:18 - 03:22
emissions of five cars over their entire lifetime.
03:22 - 03:23
Additionally,
03:23 - 03:26
a more recent study showed that training GP
03:26 - 03:30
T three with 100 and 75 billion parameters
03:30 - 03:33
resulted in 502 metric tons of carbon.
03:34 - 03:39
This is the equivalent to driving 100 and 12 gas powered cars for a year overall.
03:39 - 03:42
It was concluded that GP T three's daily carbon
03:42 - 03:45
footprint was estimated to be £50 of CO2.
03:49 - 03:53
How can A I positively impact the environment?
03:53 - 03:57
A I has the ability to analyze the vast amount of data.
03:57 - 03:59
This allows it to help mitigate climate change
03:59 - 04:02
and help societies adapt to its challenges.
04:02 - 04:04
More specifically,
04:04 - 04:07
an A I analyzes complex and evolving variables
04:07 - 04:10
within the climate system to improve models,
04:10 - 04:13
narrow uncertainties and make better predictions.
04:13 - 04:18
This can help businesses who may be affected by disruptions due to the climate
04:18 - 04:21
change plan ahead to better adapt to them.
04:22 - 04:26
A I can also help us create materials that reduce energy waste.
04:26 - 04:30
For example, they can help us develop materials that are lighter and stronger,
04:30 - 04:33
which means they consume less energy.
04:33 - 04:37
It can help us design new materials that use less resources.
04:37 - 04:41
Additionally, it can monitor energy consumption, floods, deforestation,
04:41 - 04:42
illegal fishing,
04:43 - 04:43
moni
04:44 - 04:48
identify methane leaks and opportunities for increased efficiency,
04:51 - 04:54
making A I more sustainable.
04:54 - 04:59
We've seen the numbers A I technology is using billions of gallons of water,
04:59 - 05:04
vast amounts of energy and producing over half a million pounds of Co Two per model.
05:05 - 05:10
What can we do to be proactive and stop the detrimental effect of the envi
05:10 - 05:12
environment is already feeling
05:12 - 05:17
we can ensure developers know how much electricity their computers are using
05:17 - 05:21
and how that translates into carbon emissions specifically to address this issue.
05:22 - 05:25
A group of researchers from Stanford and mcgill
05:25 - 05:28
University have developed a tracker to measure energy
05:29 - 05:32
and use carbon emissions from training A I models.
05:32 - 05:36
Additionally, Microsoft developed a technology called
05:36 - 05:39
Emissions Impact Dashboard for Azure.
05:39 - 05:43
This allows businesses to calculate their cloud's carbon footprint.
05:44 - 05:46
Companies have already started
05:47 - 05:51
and have contracts committing to use renewable energy.
05:51 - 05:58
A great example of this is Microsoft who has a long term contracts for green energy
05:58 - 06:05
and has planned to be running on 100% renewable energy by as early as next year. 2025.
06:05 - 06:06
Currently,
06:06 - 06:12
Google's data centers already get 100% of their energy from renewable resources
06:12 - 06:13
resources.
06:13 - 06:17
Companies can better maintain and manage their computers.
06:17 - 06:20
There are many ways to make computers more energy efficient which
06:20 - 06:24
in turn will save electricity along with maintaining their computer usage.
06:25 - 06:27
They need to update their technology with more
06:27 - 06:31
efficient hardware which requires less energy for cooling
06:32 - 06:34
algorithms have different strengths and weaknesses.
06:34 - 06:37
So finding the most efficient one is important.
06:37 - 06:40
It's equally as important to find the right A I model.
06:41 - 06:42
It would be irresponsible,
06:42 - 06:48
irresponsible to choose a larger model rather than a smaller model that can serves
06:48 - 06:50
resources.
06:53 - 06:56
A is relation to macroeconomics.
06:56 - 07:01
A A I's technology can enhance productivity and efficiency by
07:01 - 07:05
making predictions doing the task in place of a worker
07:05 - 07:07
and find the most efficient processes.
07:07 - 07:12
This leads to more products produced at a lower cost of resources.
07:12 - 07:16
With A I technology on the rise, it is changing the labor market.
07:16 - 07:20
For example, some jobs are being pushed out while others are being created.
07:20 - 07:22
While this is happening. It's changing
07:22 - 07:27
unemployment rates and wages which all play a part in macroeconomics.
07:28 - 07:33
A I also plays a vital role in the financial market for fraud and risk,
07:33 - 07:35
risk detection.
07:36 - 07:38
And these are my sources.
07:41 - 07:47
Um I very much appreciate you guys taking the time to watch this video.
07:48 - 07:51
Um I hope you guys learned as much as I did.
07:52 - 07:55
Um It was great talking with you guys and thank you.
See full transcript