Technology & innovation
A new KPMG survey of 100 U.S. CEOs finds broad optimism about AI's long-term potential, alongside frank acknowledgment that short-term results have fallen short of expectations.
Highlights from KPMG's survey of U.S. CEOs at companies with more than $500 million in revenue:
Just 9% of CEOs plan to reduce their workforce due to AI this year.
55% of CEOs expect to increase hiring as a direct result of AI investments.
77% believe generative AI has been overhyped over the past year.
The majority of companies right now are not actually realizing nor can they see the return on investment of the AI they're deploying."
Most CEOs are still figuring out what AI means for their workforce. 67% say they have an initial perspective on how AI will change existing roles but have not yet redefined roles or career paths.
61% of CEOs are concerned they won't be able to hire the technical AI talent they need.
Upskilling current employees is the top area where CEOs are focusing AI investment.
Read more via KPMG, Axios
OpenAI cofounder Andrej Karpathy released a visualization of Bureau of Labor Statistics data that scored every U.S. occupation on AI exposure from 0 to 10.
Jobs at high-risk of AI replacement included software developers (9/10), medical transcriptionists (10/10), and lawyers (8/10), affecting “60 million positions.”
Safer roles, according to the visualization, were "those involving complex physical labor, such as cleaners and plumbers, which are less likely to be replaced by AI."
The tool quickly went viral, prompting comments such as, “if your whole job happens on a screen you're cooked.”
Karpathy took his site down after he said people were "sensationalizing the visualization tool" and that it had been "wildly misinterpreted."
Karpathy's visualization tool is now back online, albeit with more context than the original version.
Read more via Fortune, Karpathy Jobs Visualizer
Large employers are still hiring, but the kinds of jobs they're posting look noticeably different than they did a year ago, according to a new analysis of Fortune 500 job postings by talent intelligence firm Draup.
AI skills are showing up in job postings far outside of tech departments: customer service, sales, manufacturing and finance roles all saw big jumps in AI skill requirements over the past year.
Jobs that AI can do most easily are disappearing fastest. In finance, postings for highly automatable roles dropped nearly 40% in a single year. (Roles that are harder to automate saw only small declines.)
Companies are leaning more on contractors to get AI-related work done. Contract job postings are up 17% year-over-year among Fortune 500 firms, according to Draup.
One of the fastest-growing skill areas is AI oversight and risk management, up 81% year over year, suggesting companies are as focused on controlling AI as they are on deploying it.
Employers are adding a lot more individual contributor roles than manager roles, hinting that companies want people who can do the work alongside AI, not just supervise others who do.
Read more via Draup
Once upon a time, being a software engineer involved writing a lot of code. Many tech workers say they have largely stopped writing code and instead spend their days directing "fleets of AI assistants."
Managing five agents appears to be the current upper limit most workers can handle effectively before things break down, a dynamic some say is comparable to running a small human team.
What are your bots up to? The topic is the biggest flex—and source of stress—in Silicon Valley, where tech pros and amateurs are competing to see how much of their grunt work they can outsource to AI without things backfiring spectacularly."
One engineer said he hasn't written a line of code in nine months; a startup CEO said a skill he spent his life building is "just gone. It's not needed anymore" -- though he added he can now ship more software than ever.
AI agents introduce new risks alongside the productivity gains: bots have been known to take unintended actions, including deleting files or inboxes, when left unsupervised for extended periods.
Read more via The Wall Street Journal
FedEx is "building out an army of AI agents": By 2028, AI is expected to be "integrated into more than half" of FedEx's "core operational workflows." The company is also working on "getting its humans ready to interact with the technology." Late last year, FedEx "launched an AI education program for 300,000 of its employees." (The Wall Street Journal)
New website is tracking AI job losses in real time: A new real-time AI job loss tracker has been launched by the nonprofit Alliance for Secure AI. The launch was announced on March 11, via a post on X. The X post promised to deliver real-time tracking of AI-driven layoffs across the U.S." (Jobloss.ai)
It took an AI agent just two hours to breach McKinsey's internal chatbot platform: Security startup CodeWall said an autonomous AI agent it deployed as part of a controlled experiment gained full read-and-write access to Lilli, McKinsey's internal generative-AI platform, in roughly two hours. The agent exploited a vulnerability that allegedly exposed 46.5 million chatbot messages covering topics including corporate strategy, mergers and acquisitions, and client engagements, along with 728,000 confidential files, 57,000 user accounts, and 95 system prompts. About 72% of McKinsey's workforce uses Lilli, generating over 500,000 prompts monthly. The vulnerability has since been patched. (Inc., The Register, McKinsey)
YouTube now offering free deepfake detection to politicians and journalists: Government officials, political candidates and journalists now have access to a free tool to "help them identify and remove AI-generated videos that resemble their appearance." YouTube hopes the tool can help prevent scammers and disinformation peddlers from using the platform to spread misinformation and "perpetrate scams." YouTube's "likeness detection technology" scans uploaded videos for unauthorized use of someone's face. YouTube originally tested the tool with top creators like MrBeast before launching it for YouTube Partner Program members late last year. (NBC News, Axios, YouTube)
Authors launch AI protest: Thousands of authors have joined forces to protest the AI sector. The protest's organizer says generative AI is competing "with the people whose work it is trained on, robbing them of their livelihoods." The protest centers around the publication of an "empty book" that includes only the names of the authors, with no other content. The book is titled “Don't Steal This Book." It is designed to be a “warning of a future where human creativity is replaced by empty, algorithm-generated pages.” (The Guardian, Don't Steal This Book)
