April 22, 2025 • 12 mins
In this episode of #Trending, host Jim Love discusses Google's recent legal defeat over its ad tech monopoly, with the US District Court ruling that Google violated antitrust laws. The show also examines a new paper by DeepMind researchers outlining a roadmap for AI to exceed human intelligence through experiential learning. Furthermore, it covers Microsoft and Western Digital's initiative to recycle hard drives and recover rare earth materials to combat electronic waste and supply chain dependencies. Lastly, the impact of polite language on ChatGPT's operational costs is highlighted, revealing significant financial and environmental implications.
00:00 Google's Ad Tech Monopoly Verdict
02:51 AI's Path to Surpassing Human Intelligence
07:32 Recycling Rare Earth Elements from Hard Drives
09:57 The Cost of Politeness in AI Interactions
11:42 Conclusion and Contact Information
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:02):
Google was found guilty ofhaving an ad tech monopoly.
The company's roadmap for AI toexceed human intelligence, Microsoft
and Western Digital recycle harddrives and recover rare materials,
and your parents were wrong.
It does cost you to bepolite well to an ai.
Anyway.
Welcome to hashtag Trending.
(00:23):
I'm your host, Jim Love.
Let's get into it.
A US District Court has ruled that Googleillegally monopolized key segments of
the online advertising market marking asignificant defeat for the tech giant.
Last Thursday, the court found thatGoogle violated antitrUSt laws by
willfully acquiring and maintainingmonopoly power in the markets for
(00:45):
publisher ad servers and ad exchanges.
The plaintiffs have proven thatGoogle has willfully engaged in a
series of anti-competitive acts toacquire and maintain monopoly power
in the publisher ad server and adexchange markets for open web display
advertising was the judge's conclUSion.
This decision opens the door forpotential structural remedies,
(01:07):
including the divestiture of partsof Google's advertising bUSiness.
Now, while the ruling is a major setback,the legal battle is far from over
Google plans to appeal the decision,arguing that its ad tech tools are
competitive and beneficial to publishers.
The company maintains that it earnedits market position fairly, and the
(01:28):
disrupting the ecosystem could harmcompetitors and users' privacy.
The potential consequences of the rulingare significant, and they could reshape
the digital advertising landscape andimpact how online ads are bought and sold.
But despite these challenges, Googleremains confident in its legal position.
The company argues that the marketdefinitions used in the case were too
(01:51):
narrow and that its ad tech tools operatein a highly competitive environment.
Google believes that upon appeal, itcan demonstrate that its practices
have not harmed competition orconsumers despite their optimism.
This is another defeat for Google.
Last year, the company lost a hugecase when a judge ruled that its
practice of paying other tech.
(02:13):
Companies to make it search engine.
The default on devices andbrowsers was anti-competitive.
And
Based on that ruling, the DOJ recommendedthat Google be forced to sell Chrome,
the world's most popular browser, andto decouple it from Android, which is
also the world's most popular mobile os.
The company could even potentiallybe forced to sell Android.
(02:35):
And if that's not enough, Google'sfacing similar actions in Canada, the
uk, Europe, China, India, and Japan.
The actions and the appeals will go on forsome time, but right now the score is two.
Nothing.
A new paper published last week with thetitle, the Era of Experience, paves the
(02:56):
Way for a New Wave of AI developments thatthe authors claim will ultimately lead to
AI surpassing human intelligence abilitiesby learning from its own experiences.
It was published by Deep Mind ResearchLead David Silver and Reinforcement
Learning, pioneer Richard Sutton.
The paper argues that the classic formof learning and training of generative
(03:19):
AI models ingesting huge amounts ofpreexisting data could actually be
holding AI back To some extent, existingAI model builders already know this
and they've been introducing what theycall reinforcement learning, where
humans interact with the AI model tolet it learn by giving it rewards,
for finding strategies that solveproblems or execute tasks successfully.
(03:44):
This method has led to hugeincreases in AI performance,
Silver and Sutton argue that USing thismethod and this method alone, no matter
how sophisticated makes AI an echochamber of existing human knowledge.
In other words, although AI hasmade exceptional progress, it
ultimately reaches a limit imposedby the data we give it and the
(04:07):
ability of its human trainers.
In order to grow beyond those limitations,the authors say, AI agents must
actively engage with the world, collectobservational data, and then use that data
to iteratively refine their understanding,mirroring in many ways the process
that drives human scientific progress.
(04:27):
And it's not farfetched.
It builds on the accomplishment of theprogram Alpha Go, which eventually beat
the masters in the strategy game go.
That has limitless possibilities ofmoves, And it did it not by being given
the rules, but by trying millions andmillions of simulations and learning
new techniques that surprise and evenbeat the grand masters of the game.
(04:51):
But the authors believe that thissimulation method is also limited.
And what is needed are the actualinteractions with the real world
and learning, not task by task,but by being given larger goals and
learning the steps to achieve them.
With the advent of what's been calledagents that can actually operate computers
from human interfaces, this allows AI tohave real experiences that it can learn
(05:18):
from now instead of only simulated games.
These agents can perform actions,observe true outcomes, and
adapt over months or even years.
And the method differs because it combinesthe classic reinforcement learning
concepts like value functions, andworld models with what the author's term
grounded rewards drawn from real data.
(05:40):
Things like health metrics,sensor ratings, or even
scientific experiment results.
In practice, an agent could refine its ownhypotheses by running lab tests, tweaking
parameters, and measuring the impact, andthen repeating that cycle autonomoUSly.
And by continually generating andlearning from its own experiential data,
(06:00):
far beyond any static human data set.
The authors believe AI can not onlymatch, but will actually surpass human
expertise even in complex domains.
And for those who worry about theseautonomoUS intelligent agents being
dangeroUS, the authors acknowledgethat the real world experimentation
(06:22):
does carry some risk, but conversely,might actually make AI safer.
There's a famoUS example that showsthe danger of ai where an AI was
given a task of being exceptionallyefficient at making paperclips.
In this, the researchers realizedthat left to its own devices, the AI
might even destroy the world becauseof its singular goal of turning the
(06:44):
world's resources into paperclips.
But with this new approach, the agents canbe given larger goals and be tasked with
monitoring their real consequences andadjUSting misaligned objectives over time.
The paper has garnered some attention,but it's largely flying under the radar,
even though it's the first roadmap tobreaking the constructs of existing AI
(07:08):
models and surpassing human intelligenceand abilities, and may one day rank
with Google's other monumental paper.
The 2017 paper Attention is AllYou Need, which introduced the
transformer model revolutionizednatural language processing and led
to the development of the AI models.
Like ChatGPT and others.
(07:32):
Lately, most of US have become aware ofsomething called rare earth minerals,
something we'd never heard of before,but we found out these are hugely
important in our computers and our lives.
China has very recently started torestrict their export as part of
an ongoing tariff war with the US.
As we've pointed out in earlier stories,these rare earths are essential in
(07:53):
a number of areas and restrictingthem could turn into a real crisis.
So the US has been floating variousstrategies to get rare earth deposits
from Ukraine or even Greenland and Canada,which also have huge deposits, but so far.
Those negotiations have been, to putit kindly, not very successful, and
(08:13):
realistically, it could take years toreplace not jUSt the Chinese supplies,
but there's also another piece that'sthe refining capability, which has
incredible environmental impacts, andthat's one that China also provides.
So Western Digital and Microsoft haveteamed up to launch a new recycling
initiative aimed at extracting these rareearth elements from old hard disc drives.
(08:37):
This has a dual benefit.
It provides a source of theserare materials, but it also
reduces electronic waste whileenhancing domestic supply chains.
the program announced in mid-Aprilwill use demagnetization technology to
remove data from end of life enterprisehard disc drives while preserving
the rare earth magnets inside.
And these magnets typically verydifficult to recover, will then be
(09:00):
extracted and reused in new hard drivesor even other electronic products.
Microsoft plans to implement the systemacross its cloud data centers as part of
a. Broader push towards sustainabilityand circular hardware practices.
The recovered magnets will beused in servers, wind turbines,
and even electric vehicles.
the initiative addresses two growingconcerns, the environmental toll of
(09:24):
e-waste, and the dependency on importedrare earths, much of which come from
restricted supplies in China and byrecovering magnets and other materials
domestically, Microsoft and WesternDigital aim to reduce both landfill volume
and geopolitical supply chain risks.
The pilot program could become a model forother tech giants with large scale data
(09:46):
infrastructures, and as an added bonus,this also fits into Microsoft's climate
commitments, which include becomingcarbon negative and zero waste by 2030.
turns out your parents were wrong.
Politeness has a price.
In fact, every time you typeplease or thank you into ChatGPT,
(10:06):
you're contributing to a growingoperational cost for ChatGPT.
CEO Sam Altman recently revealed thatthese polite exchanges are costing the
company tens of millions of dollarsannually in electricity expenses.
And While these courtesies makeinteractions feel more human, they also
add to the computational load impactingboth finances and the environment.
(10:30):
ChatGPT processes over a billionrequests daily from approximately
350 million weekly active users.
Each additional word, including politephrases, increases the number of tokens
the AI must process leading to higherenergy consumption and costs, And
Altman acknowledges this expense, butdeemed it tens of millions of dollars
(10:52):
well spent emphasizing the valueof human-like interactions and the
environmental impact is also significant.
Generating a single response canconsume about 2.9 watt hours of
electricity, and when this is scaledacross millions of interactions, the
energy usage becomes substantial.
Despite this many users continue touse polite language with a 2024 survey
(11:17):
indicating that 67% of Americans doso and I assume a hundred percent of
Canadians driven by habit or desire tomaintain good relationships with the AI
So while your polite interactions withChatGPT contribute to a more human-like
experience, they have real world costs.
And geez, I'm Canadian.
(11:38):
I don't even want to knowwhat it costs to say, sorry.
And that's our show.
I'd say thanks for listening, butwe don't have a big budget, so I'll
just say it's good to be back and itdoesn't cost a thing to talk to us
you can reach me ateditorial@technewsday.ca or on
LinkedIn or in the comments on YouTube.
(11:59):
I'm your host, Jim Love.
Have a terrific Tuesday.
Google was found guilty ofhaving an ad tech monopoly.
The company's roadmap for AI toexceed human intelligence, Microsoft
and Western Digital recycle harddrives and recover rare materials,
and your parents were wrong.
It does cost you to bepolite well to an ai.
Anyway.
Welcome to hashtag Trending.
(00:23):
I'm your host, Jim Love.
Let's get into it.
A US District Court has ruled that Googleillegally monopolized key segments of
the online advertising market marking asignificant defeat for the tech giant.
Last Thursday, the court found thatGoogle violated antitrUSt laws by
willfully acquiring and maintainingmonopoly power in the markets for
(00:45):
publisher ad servers and ad exchanges.
The plaintiffs have proven thatGoogle has willfully engaged in a
series of anti-competitive acts toacquire and maintain monopoly power
in the publisher ad server and adexchange markets for open web display
advertising was the judge's conclUSion.
This decision opens the door forpotential structural remedies,
(01:07):
including the divestiture of partsof Google's advertising bUSiness.
Now, while the ruling is a major setback,the legal battle is far from over
Google plans to appeal the decision,arguing that its ad tech tools are
competitive and beneficial to publishers.
The company maintains that it earnedits market position fairly, and the
(01:28):
disrupting the ecosystem could harmcompetitors and users' privacy.
The potential consequences of the rulingare significant, and they could reshape
the digital advertising landscape andimpact how online ads are bought and sold.
But despite these challenges, Googleremains confident in its legal position.
The company argues that the marketdefinitions used in the case were too
(01:51):
narrow and that its ad tech tools operatein a highly competitive environment.
Google believes that upon appeal, itcan demonstrate that its practices
have not harmed competition orconsumers despite their optimism.
This is another defeat for Google.
Last year, the company lost a hugecase when a judge ruled that its
practice of paying other tech.
(02:13):
Companies to make it search engine.
The default on devices andbrowsers was anti-competitive.
And
Based on that ruling, the DOJ recommendedthat Google be forced to sell Chrome,
the world's most popular browser, andto decouple it from Android, which is
also the world's most popular mobile os.
The company could even potentiallybe forced to sell Android.
(02:35):
And if that's not enough, Google'sfacing similar actions in Canada, the
uk, Europe, China, India, and Japan.
The actions and the appeals will go on forsome time, but right now the score is two.
Nothing.
A new paper published last week with thetitle, the Era of Experience, paves the
(02:56):
Way for a New Wave of AI developments thatthe authors claim will ultimately lead to
AI surpassing human intelligence abilitiesby learning from its own experiences.
It was published by Deep Mind ResearchLead David Silver and Reinforcement
Learning, pioneer Richard Sutton.
The paper argues that the classic formof learning and training of generative
(03:19):
AI models ingesting huge amounts ofpreexisting data could actually be
holding AI back To some extent, existingAI model builders already know this
and they've been introducing what theycall reinforcement learning, where
humans interact with the AI model tolet it learn by giving it rewards,
for finding strategies that solveproblems or execute tasks successfully.
(03:44):
This method has led to hugeincreases in AI performance,
Silver and Sutton argue that USing thismethod and this method alone, no matter
how sophisticated makes AI an echochamber of existing human knowledge.
In other words, although AI hasmade exceptional progress, it
ultimately reaches a limit imposedby the data we give it and the
(04:07):
ability of its human trainers.
In order to grow beyond those limitations,the authors say, AI agents must
actively engage with the world, collectobservational data, and then use that data
to iteratively refine their understanding,mirroring in many ways the process
that drives human scientific progress.
(04:27):
And it's not farfetched.
It builds on the accomplishment of theprogram Alpha Go, which eventually beat
the masters in the strategy game go.
That has limitless possibilities ofmoves, And it did it not by being given
the rules, but by trying millions andmillions of simulations and learning
new techniques that surprise and evenbeat the grand masters of the game.
(04:51):
But the authors believe that thissimulation method is also limited.
And what is needed are the actualinteractions with the real world
and learning, not task by task,but by being given larger goals and
learning the steps to achieve them.
With the advent of what's been calledagents that can actually operate computers
from human interfaces, this allows AI tohave real experiences that it can learn
(05:18):
from now instead of only simulated games.
These agents can perform actions,observe true outcomes, and
adapt over months or even years.
And the method differs because it combinesthe classic reinforcement learning
concepts like value functions, andworld models with what the author's term
grounded rewards drawn from real data.
(05:40):
Things like health metrics,sensor ratings, or even
scientific experiment results.
In practice, an agent could refine its ownhypotheses by running lab tests, tweaking
parameters, and measuring the impact, andthen repeating that cycle autonomoUSly.
And by continually generating andlearning from its own experiential data,
(06:00):
far beyond any static human data set.
The authors believe AI can not onlymatch, but will actually surpass human
expertise even in complex domains.
And for those who worry about theseautonomoUS intelligent agents being
dangeroUS, the authors acknowledgethat the real world experimentation
(06:22):
does carry some risk, but conversely,might actually make AI safer.
There's a famoUS example that showsthe danger of ai where an AI was
given a task of being exceptionallyefficient at making paperclips.
In this, the researchers realizedthat left to its own devices, the AI
might even destroy the world becauseof its singular goal of turning the
(06:44):
world's resources into paperclips.
But with this new approach, the agents canbe given larger goals and be tasked with
monitoring their real consequences andadjUSting misaligned objectives over time.
The paper has garnered some attention,but it's largely flying under the radar,
even though it's the first roadmap tobreaking the constructs of existing AI
(07:08):
models and surpassing human intelligenceand abilities, and may one day rank
with Google's other monumental paper.
The 2017 paper Attention is AllYou Need, which introduced the
transformer model revolutionizednatural language processing and led
to the development of the AI models.
Like ChatGPT and others.
(07:32):
Lately, most of US have become aware ofsomething called rare earth minerals,
something we'd never heard of before,but we found out these are hugely
important in our computers and our lives.
China has very recently started torestrict their export as part of
an ongoing tariff war with the US.
As we've pointed out in earlier stories,these rare earths are essential in
(07:53):
a number of areas and restrictingthem could turn into a real crisis.
So the US has been floating variousstrategies to get rare earth deposits
from Ukraine or even Greenland and Canada,which also have huge deposits, but so far.
Those negotiations have been, to putit kindly, not very successful, and
(08:13):
realistically, it could take years toreplace not jUSt the Chinese supplies,
but there's also another piece that'sthe refining capability, which has
incredible environmental impacts, andthat's one that China also provides.
So Western Digital and Microsoft haveteamed up to launch a new recycling
initiative aimed at extracting these rareearth elements from old hard disc drives.
(08:37):
This has a dual benefit.
It provides a source of theserare materials, but it also
reduces electronic waste whileenhancing domestic supply chains.
the program announced in mid-Aprilwill use demagnetization technology to
remove data from end of life enterprisehard disc drives while preserving
the rare earth magnets inside.
And these magnets typically verydifficult to recover, will then be
(09:00):
extracted and reused in new hard drivesor even other electronic products.
Microsoft plans to implement the systemacross its cloud data centers as part of
a. Broader push towards sustainabilityand circular hardware practices.
The recovered magnets will beused in servers, wind turbines,
and even electric vehicles.
the initiative addresses two growingconcerns, the environmental toll of
(09:24):
e-waste, and the dependency on importedrare earths, much of which come from
restricted supplies in China and byrecovering magnets and other materials
domestically, Microsoft and WesternDigital aim to reduce both landfill volume
and geopolitical supply chain risks.
The pilot program could become a model forother tech giants with large scale data
(09:46):
infrastructures, and as an added bonus,this also fits into Microsoft's climate
commitments, which include becomingcarbon negative and zero waste by 2030.
turns out your parents were wrong.
Politeness has a price.
In fact, every time you typeplease or thank you into ChatGPT,
(10:06):
you're contributing to a growingoperational cost for ChatGPT.
CEO Sam Altman recently revealed thatthese polite exchanges are costing the
company tens of millions of dollarsannually in electricity expenses.
And While these courtesies makeinteractions feel more human, they also
add to the computational load impactingboth finances and the environment.
(10:30):
ChatGPT processes over a billionrequests daily from approximately
350 million weekly active users.
Each additional word, including politephrases, increases the number of tokens
the AI must process leading to higherenergy consumption and costs, And
Altman acknowledges this expense, butdeemed it tens of millions of dollars
(10:52):
well spent emphasizing the valueof human-like interactions and the
environmental impact is also significant.
Generating a single response canconsume about 2.9 watt hours of
electricity, and when this is scaledacross millions of interactions, the
energy usage becomes substantial.
Despite this many users continue touse polite language with a 2024 survey
(11:17):
indicating that 67% of Americans doso and I assume a hundred percent of
Canadians driven by habit or desire tomaintain good relationships with the AI
So while your polite interactions withChatGPT contribute to a more human-like
experience, they have real world costs.
And geez, I'm Canadian.
(11:38):
I don't even want to knowwhat it costs to say, sorry.
And that's our show.
I'd say thanks for listening, butwe don't have a big budget, so I'll
just say it's good to be back and itdoesn't cost a thing to talk to us
you can reach me ateditorial@technewsday.ca or on
LinkedIn or in the comments on YouTube.
(11:59):
I'm your host, Jim Love.
Have a terrific Tuesday.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!