Microsoft goes nuclear, an Atlassian study debunks the idea that the return to office has better results, there’s a big gap between what AI can to in terms of eliminating jobs and what workers fear it can do – at least for now, deep fakes hit the New Hampshire primary and if AI kills the internet, it might be a murder suicide.
 |
 |
 |
All this and more on this edition of Hashtag Trending. I’m your host Jim Love, CIO of IT World Canada and TechNewsDay in the US.
Microsoft is accelerating efforts to power data centers with nuclear energy. The tech giant just hired Erin Henderson, a former nuclear plant executive, to lead the push.
Henderson spent 13 years at the Tennessee Valley Authority working on major nuclear sites. Now she’ll drive Microsoft’s small modular reactor and microreactor strategy.
Demand for data centers is booming, but grids are struggling to provide enough clean energy. So Microsoft is getting creative, even teaming up to train AI on nuclear regulations.
The company has bought credits from Canadian utility OPG’s nuclear plants. And it signed a 24/7 deal with Constellation for a Virginia data center.
Nuclear saw support at last year’s major climate summit COP28. But the industry also faces turbulence – a leading small modular reactor firm had a major project canceled recently.
Microsoft appears undeterred, charging ahead to develop next-gen nuclear. With grids maxing out, data centers need solutions. And companies like Microsoft have the scale to spark innovation.
Sources include: Data Center Dynamics
There’s been a number of stories of CEOs pushing employees back to the office, but is that working?
Despite the stories of employees being forced back, flexible work is gaining more acceptance, even with executives who have resisted it. One study noted that remote versus in office work numbers have stabilized at about a 50/50 split. But what works better?
Atlassian, a company that has embraced remote work set out to find out the answer. They surveyed 5000 of their own employees and also 100 Fortune 1000 and 100 Fortune 500 CEOs. The key insight: 92 per cent of Atlassian’s staff say flexibility is critical for their best work.
And 99 per cent of CEOs they surveyed now believe distributed work is the future.
1 in 3 executives say mandatory office policies haven’t impacted productivity at their firms. They increasingly realize you can focus on how work happens, not where.
And at Atlassian 91 per cent of the staff say that the ability to have remote work is one of the main reasons they stay with the company.
Atlassian’s head of remote work research Annie Dean isn’t surprised by the data. Past evidence shows forced commuting damages morale and retention. Workers want autonomy in location.
Dean says hybrid models just create an “illusion of choice” with their mandatory days. Creativity depends on management, not physical proximity. Maybe we’re finally moving past tired water cooler myths.
The message from employees and CEOs is clear: help people succeed on their own terms. Trust and flexibility now look essential, even to formerly strict executives. Work can thrive whether you’re at home or HQ.
Sources include: CNN
To paraphrase Mark Twain, rumours of the death of our jobs may be somewhat exaggerated.
A survey by a firm called Spokeo found 66 per cent of U.S. workers think AI could perform their jobs.
Nearly 75 per cent are concerned about AI’s impact on their industry.
But there’s optimism too – over 75 per cent believe AI will reduce workplace stress and shorten the workweek.
Still, people see the threat to incomes in the near term. So that’s what this group of U.S. workers think.
But how much can AI actually replace?
An MIT study dug into the actual feasibility of AI automation. It modeled the economics of visual AI across 800 occupations.
The finding?
This study found that only 23 per cent of U.S. wages are in roles where machine vision automation would be cost effective today. It would require major tech cost reductions to boost viability. Why machine vision? An analysis of jobs like bakery workers showed humans still do key visual tasks cheaper than deploying cameras and AI.
Not that some jobs won’t be affected. AI replacements,areas like retail and healthcare show more potential.
So while new AI tools create anxiety about jobs, the reality is automation has limits today on both tech capability and budgets.
Worker surveys reveal high anxiety over being replaced by thinking machines. But detailed economic analysis finds most roles aren’t ripe for cost-effective automation yet.
At the speed AI is developing, who knows what the answer will be a year from now. But today, humans still have an edge at most tasks due to the high hidden costs of AI systems.
Sources include: National Post and FoxBusiness
_____
It’s started – a deepfake with Biden’s voice is used to subvert New Hampshire primary voting.
New Hampshire’s presidential primary got hit with a concerning dirty trick this weekend. Voters received a robocall in Joe Biden’s voice, urging Democrats not to vote on Tuesday. It said to save your vote for November instead.
The call said it was from the phone number of prominent Democrat Kathy Sullivan. She’s running a super PAC to organize a Biden write-in campaign, since he’s not on the official ballot per DNC rules.
Sullivan and state officials blasted the deception and apparent attempt at voter suppression. The state Attorney General is now investigating potential violations of law.
The use of what sounds like artificially generated audio impersonating Biden raises alarms about “deep fake” techniques spreading to campaigns. Sullivan wants the source prosecuted fully.
Biden’s team is exploring further actions in response. Senator Maggie Hassan hopes it instead drives higher turnout to back Biden. But no matter the impact, the dark political dirty trick signals new ethical issues as AI capabilities advance.
All of this happens on the day that 11 Labs, one of the best known firms in AI voice production officially became a unicorn – worth more than 1 billion dollars.
Something tells me the genie isn’t going back into the bottle on this one.
Sources include: NBC News and Reuters
And if video killed the radio star, will AI kill the internet? Or will it be a suicide?
It’s called model decay…
Researchers are warning that AI models could poison themselves by training on other AIs’ synthetic output. These models, at least for now, need vast amounts of data. And it turns out that they are training on data created by other AI models – what some have termed the “slime” that is killing the internet.
Some examples of this model decay? Elon Musk’s bot Grok plagiarized an OpenAI response, exposing how it had “learned” from OpenAI code.
Amazon product listings are turning up with oddly named products called “OpenAI policy errors.”
The next step in model decay is model collapse. As AIs scrape more web data, including each other’s made up text, they lose touch with reality. Output quality declines.
Many developers still deny these risks, but are they hesitant to admit flaws in promising new revenue generators? With models using unfathomable training data, there may be no reversing the damage.
One academic says we’re past hoping AIs can tell human versus bot content apart. The models will keep soaking up more machine falsehoods. She warns the internet itself may burn from disinformation.
In the end, last year’s AI promise may this year become an unreliability nightmare. The tech could spread spam and lies faster than we can correct. And that may only accelerate as adoption grows.
But before we predict the end of the world there are a few other data points to consider. Sam Altman talked last year about being able to use “synthetic data” to train and AI, making them less reliant on having to do vast scraping of Internet data. OpenAI and others are courting reliable sources like news outlets to find more accurate info. And I’m doing a story this week in IT World Canada on how what is termed RAG – short for Retrieval Augmented Generation where companies can use the conversational ability of AI engines, but restrict the answers to well contained and accurate data sets.
Watch this space.
Sources include: Analytics India
Hashtag Trending goes to air five days a week with a daily news show and every Saturday, we have an interview show called the Weekend Edition.
We love your comments. Please let us know what you think. You can reach me at [email protected] or leave a comment under the show notes at www.itworldcanada.com/podcasts.
I’m your host Jim Love, thanks for listening and have a Terrific Tuesday.
