The US is widening the innovation gap with Europe when it comes to leveraging AI. Microsoft is pulling out the stops to convince government clients that they’ve changed their security culture. GPT outperforms financial analysts. A Starbucks in Korea is run by robots.And if tape is really dead, why are we still shipping so much of it?
These stories and more on the “slow and steady doesn’t always win the race “ edition of Hashtag Trending. I’m your host Jim Love, let’s get into it.
Economists are warning of a potential widening innovation and economic gap between the U.S. and Europe when it comes to capturing the benefits of generative AI. A new paper by the McKinsey Global Institute finds that most AI breakthroughs are happening in the U.S., while Europe moves more cautiously in regulating labor-displacing technologies.
European Central Bank president Christine Lagarde, addressing the Council on Foreign Relations last month said, “It’s just mind boggling that productivity growth in the United State between 2019 and now has been 6%. In Europe, its .6%.”
Canadian listeners should not be smug about this. Canadian productivity, according to an article published in the Bank of Montreal economic blog, has been declining by negative .3 percent per year.
The paper points to structural factors like aging workforces that could stall workforce transitions required for widespread AI adoption in Europe. It estimates both regions may need 12 million workers to shift roles in just 5 years – double the pre-pandemic pace for Europe.
Anu Madgavkar, a McKinsey partner said, “In the U.S., this rate of labor transition has been higher historically. It would require a step up in Europe’s metabolic rate of people moving to new roles.”
While rapid U.S. adoption could accelerate productivity growth from historical norms, slower European AI uptake may limit gains to just 0.3% through 2030, widening the technology and economic divide.
Sources: Axios and BMO Economics blog
Cybersecurity researchers are sounding the alarm on sophisticated phishing campaigns that are abusing Cloudflare Workers, a development and deployment infrastructure from security company Cloudflare, to serve fake login pages and harvest victims’ credentials. This “transparent phishing” or “adversary-in-the-middle” technique acts as a reverse proxy, intercepting traffic between the user and legitimate sites.
Jan Michael Alcantara of Netskope stated, “The attacker collects its web request metadata. Once the victim enters credentials, they’ll be logged in legitimately, but the attacker collects the tokens and cookies.”
Campaigns are using an evasion method called “HTML smuggling” which allows attackers to serve fraudulent HTML pages which don’t raise and red flags with anti-malware software.
Phishers are employing techniques to iframe malicious sites into genuine Microsoft HTML pages and more
Generative AI is also being leveraged to craft more effective phishing lures, including fake invoices and malicious login pages for fraudulent PDF downloads.
Many of the new attack tools can be obtained as “phishing as a service” toolkits which make adopting sophisticated techniques available to even relatively unsophisticated attackers or that save development time for even more sophisticated players.
The bottom line is that attackers are constantly evolving their phishing playbook, compromising legitimate services and employing new obfuscation and evasion strategies to bypass traditional defenses and deliver malware payloads effectively.
Sources include: The Hacker News
In a study from the University of Chicago’s Booth School of Business, researchers found that GPT-4 can outperform human analysts in the complex task of financial statement analysis. The study tasked GPT-4 with predicting future earnings directions based on standardized financial data using a “Chain of Thought” reasoning process that mimics human analysts’ steps.
At 53 percent or less, the professional analysts were performing at about the same rate as a coinflip. Remarkably, GPT-4, using the Chain of Thought process, achieved 60.35% accuracy, a full 7 percentage points higher than professional analyst forecasts made one month after earnings releases.
As one researcher stated:
“Financial statement analysis requires broad knowledge, common sense and narrative context like industry trends – things machines traditionally struggle with. Yet GPT-4 matched human-level performance, which is extraordinary.”
The researchers employed two methods – a simple prompt and a more guided “Chain of Thought” process walkthrough. GPT-4 outperformed both human analysts and specialized AI models using the latter approach.
Researchers theorize GPT-4 gleaned insights from financial ratios and trends, combining this quantitative analysis with its broad economic knowledge and theoretical underpinnings.
The findings highlight GPT-4’s remarkable general capabilities and potential to augment or replace human expertise in complex, open-ended analytical tasks well beyond its narrow training domain. As AI models continue advancing, their adoption across knowledge worker roles may accelerate rapidly.
Sources include: Medianama.com
Microsoft is pulling out all the stops to actively engage with federal government IT and cybersecurity leaders to outline its revamped internal security practices and rebuild trust following last summer’s damaging China-linked email hack that exposed top officials’ communications.
The tech giant is briefing federal agencies on its new “Secure Future Initiative” strategy that incentivizes prioritizing cybersecurity in product design and development. Microsoft executives like Steve Faehl have been meeting agency CIOs, CISOs and security teams.
The briefings come ahead of Microsoft President Brad Smith’s testimony next month before the House Homeland Security Committee regarding the hack, which a government advisory board labeled “preventable.”
In these agency meetings, a key focus has been assuring skeptics that Microsoft’s security overhaul is genuine. As Faehl stated, “They understand norms have changed…Getting certainty around our efforts is the focus.”
The outreach underscores how last year’s incident harmed government trust in Microsoft and provided an opening for rivals like Google, which recently offered discounts to lure away customers. However, Microsoft is not treating these as competitive threats, at least not in their public statements.
As Faehl put it: “We don’t see companies with alternative solutions as the competition…We are looking for partnerships to address those threats” targeting both Microsoft and the government.
Sources include: Axios
A South Korean tech giant NAVER has deployed around 100 autonomous service robots named “Rookie” to serve customers at its building’s Starbucks cafe.
In their 1784 Tower, billed as the “world’s largest robotics testbed”, you can see NAVER’s vision of a “tech convergence building” that seamlessly blends cutting-edge technologies like robotics, AI, and cloud computing into the workspace.
On a regular day, a wheeled robot called a “Rookie” autonomously navigate all 36 floors, delivering food, drinks and parcels. It uses dedicated robot elevators. It’s assisted another type of robot called Ambidex, a dual-armed precision robot that safely handles customer interactions.
According to NAVER, “In 1784, daily service becomes daily testing, and each day’s issues become the next day’s updates” as the robots continuously learn and adapt.
Underpinning this army of robots is NAVER’s pioneering ARC (AI, Robotics, Cloud) platform that provides cloud-powered intelligence. ARC enables ultra-low latency multi-robot coordination, precise indoor navigation, task planning and more through advanced algorithms and 5G connectivity.
As Sang Hyuk Son, NAVER LABS CEO, explained: “By shifting heavy computational workloads to the cloud, our robots become a lightweight endpoint that can scale intelligence as needed while benefiting from lower costs, higher performance, and enhanced mobility.”
The only question that remains is that if robots run the Starbucks and runs the office buildings, and then AI handles the analysis and reporting work, who exactly is it that will be there to order the latte?
Sources: Interesting Engineering
People that have said that tape is a dead storage medium may have to eat their words. Tape saw a huge surge in demand last year, with manufacturers behind the LTO (Linear Tape-Open) standard shipping a staggering 152.9 exabytes worth of tape cartridges.
An exabyte is one billion gigabytes. It’s about the amount of data on 25 billion smart phones.
Key vendors in this area, HPE, IBM, and Quantum attributed this 3.14% year-over-year growth to the “rapid data generation” needs of hyperscalers, enterprises, and crucially, AI workloads which can exponentially increase unstructured data storage requirements.
As one of the vendors noted, “AI workloads can cause increases in storage requirements and costs.” Tape’s massive capacity, offline protection against ransomware, and cost-effectiveness for infrequently accessed data make it an attractive archival solution.
Even a conservative 50 EB (50,000 petabytes) of tape shipped highlights tape’s enduring relevance.
As one veteran IT architect, commented: “Tape was my first love in storage, and it’s great to see it still kicking, putting all those ‘Tape is dead’ claims to bed. For cold data archiving at hyperscale, you just can’t beat that capacity-per-dollar.”
So while tapes may never regain their ’90s heyday, this unlikely resurgence proves legacy technologies can find new life answering modern data challenges, defying premature predictions of their demise.
Sources include: The Register
And that’s it for today’s show. Remember that you can get us on Apple, Spotify or wherever you get your podcasts. We’re available on YouTube in both audio and video format.
Show notes are on Tech Newsday dot com or dot ca. Take your pick.
I’m your host Jim Love, have a wonderful Wednesday.
