Headlines this week - May 18, 2025
A look at how capital is being deployed across future opportunities
This week in the future:
1 - World’s first personalized Crispr (gene-editing) treatment saves the life of a baby. The ten-months old baby was suffering a mortal genetic disease, and he has apparently recovered after being administered three doses of a gene-editing treatment to mend the mutation that was causing the problem. According to doctors, it is still too soon to talk about “curation”, and it is also not clear whether such bespoke therapies can be widely applied (e.g. to other diseases). Still, it’s a start
2 - South Korea becoming a global leader in nuclear power technologies, including SMRs
The country has the best large-scale atomic-energy industry outside China and Russia. Bloomberg reports this week that they have been developing their own atomic technology for decades, as a response to their own energy needs and the global priority to shift away from fossil fuels.
Now, with a global nuclear energy revival under way, they are in a good position to capture a big opportunity as technology suppliers. There are plans to triple global nuclear energy production for 2050, largely under pressure from growing demand driven by data centers / technology companies. South Korean companies have plans to export 10 reactors by the end of this decade, including a recently delivered $20bn project in the Emirates
The next step will be innovative Small Modular Reactors (SMRs). The country is expecting to finalize the design of its first SMR this year, and to be able to deploy it by the mid-2030s. The vision behind SMRs is to reduce costs and times to deploy by standardizing designs in a way that makes it possible to build the components in factories and then assemble them on site. The concept remains largely untested, but there are a few of these reactors in service in China and Russia
3 -Money keeps flowing to innovation projects for electric batteries
CATL will have a (large) Hong Kong IPO to fund its global expansion. CATL, the world’s leading producer of batteries for electric cars, is planning its Hong Kong IPO, the largest so far this year, in which they expect to raise more than $4bn. This will help the company fund its global expansion, e.g. with several factories planned in Europe, including one in Zaragoza (Spain). Investors are interested in funding these projects, with batteries perceived as a key bottleneck (or driver) for the expansion of electric vehicles
A new breakthrough by GM shows that battery technology is evolving fast. We have already reported here about massive improvements by Chinese vendors in charging times and ranges. This week we learned about a General Motors’ breakthrough, the lithium manganese-rich (LMR) prismatic battery cells, which has 33% more energy density (energy / kg) than current conventional batteries and offers similar ranges at a significantly lower cost. GM’s target with this are electric trucks
4 - Positive signs for AI infrastructure demand:
Nvidia’s stock rises as concerns on demand for GPUs decrease. The share price is on track to its best month in a year (+44% from an April low). In particular, there has been a very positive reaction this week to Trump’s agreements in the Middle East (see next point) and to the recently approved changes in the chip export rules. On top of that, bullish signals from Big Tech companies about their AI CapEx are also helping
Saudi Arabia has launched Humain, a new AI infrastructure company, that will be owned by the country’s sovereign fund (PIF) and chaired by Crown Prince Mohammed bin Salman. The company will focus on developing AI technologies and infrastructure (i.e. data centers), using the most advanced computing equipment (i.e. Nvidia GPUs). It was announced a day before Trump’s visit to the area. No figures have been shared, but everyone thinks they will be massive. This will add to previously announced AI startups, like Alat, also owned by PIF, with plans to invest $100bn
India wants to be an AI infrastructure hub in SE Asia, with Airtel (the incumbent telecom operator) leading the effort. Airtel has just announced a plan to double the capacity of its data center business in less than 3 years, with an estimated investment of $600m. Longer-term, the company wants to turn into the leading data center company in the area, over the next 5-7 years. For this they will take advantage of the massive internal demand, with the government now pushing for all domestic data to be stored locally, and of the pool of highly skilled talent in the country
OpenAI makes Coreweave (AI Cloud) great again. Coreweave, the AI infrastructure specialist that recently had its IPO, was falling below initial expectations, amid plenty of concerns from investors about the evolution of demand. This week, a large contract with OpenAI was announced, with an estimated value of $4bn, to be paid in cash during the next 4 years. The stock has reacted positively
SoftBank, on its FY2024/25 earnings call, confirmed its commitment to invest $100bn in the Stargate project to build AI data centers in the US, together with OpenAI and Oracle. This had previously been put into question in a Bloomberg article that claimed that the company had not yet reached any kind of agreements with potential lenders and investors, although discussions had already been opened
5 - Can agents trigger a (risky) “intelligence explosion”? There is no consensus about this.
The race continues for AI labs to launch increasingly powerful “coding AI agents”. This week we’ve seen how OpenAI launches a new agent, “Codex”, specialized in generating code from natural language. The company is positioning the product as “a virtual co-worker for engineers”. Also Google’s DeepMind research division presented AlphaEvolve, a new agent that aims to create code to tackle big problems in math and science, and which incorporates an “automatic evaluation” system to reduce hallucinations and increase accuracy
Some analysts believe that this could lead to an “intelligence explosion” (and significant safety risks). This is the mechanism described e.g. in the recent AI2027 report, which assumes that at some point AI agents will be able to self-improve (by creating code that makes themselves more powerful) and that this will lead to (out of control) “Artificial Super Intelligence” (and all the challenges implicit in that)
However, there is no consensus about this. A new paper published this week analyzes recent algorithmic improvements in AI and reaches the conclusion that the ones with the most impact were “computing-dependent”, i.e. only produced significant advantages when deployed on large-scale models with large computing needs. According to this, access to massive amounts of computing would be key to make disruptive progress in AI. So, controlling or restricting access to computing resources could be a way to reduce safety risks. By the way, this would also validate the US current approach to restrict rivals’ access to chips, as a way to reduce access to more computing, and decelerate AI innovation
6 - Meta’s delays in its flagship LLM increase concerns that AI scaling laws may be hitting a wall
This week Meta announced a delay in the rollout of their flagship AI model, “Behemoth”. The release was initially expected for April 2025, then shifted to June, and now it has been pushed to the fall or later. This is creating new concerns for investors about how the company is spending its CapEx ($72bn planned only this year). Internally, top Meta executives are not happy with their engineers’ performance, and organizational changes in the AI product group are now being rumored
This is seen as a sign that “raw” scaling is not working any more, to increase performance. On the other hand, many analysts interpret the news as a new sign that the “standard” approach at AI labs to increase models’ performance just by expanding their scale (training data, model parameters) is running out of steam. Actually, the problem at Meta is somewhat parallel to the (less explicit) delays that are affecting the release of OpenAI’s GPT-5
If confirmed, this could open a time of uncertainty for the AI industry, even if there is still plenty of room for progress through test-time compute (e.g. with “reasoning” models) and, of course, through applications / agents targeting real-world problems, that could run on top of the existing LLMs
7 - Microsoft: lots of uncertainties after its “divorce” from OpenAI.
A long Bloomberg article this week discusses how Microsoft still has to define its role in the emerging AI landscape, after his progressive separation (still under way) from OpenAI. The company has opened to other Foundational Models beyond GPT, to defend its position as a neutral infrastructure provider. At the same time, they’ve seen how OpenAI is launching AI products that directly compete with Microsoft’s (e.g.: Copilot, which used to be Microsoft’s “flagship” AI product is now much less popular than ChatGPT itself).
This is more pressure for S Nadella, specially with investors now asking about the returns of the company’s previous investment ($13.75bn) in OpenAI. A related issue is the decision (now under discussion) about the stake that Microsoft will have in the new For Profit branch of OpenAI
8 - Are jobs more resilient (to AI) than we initially expected?
According to Box’s CEO, AI will be used to do more with the same jobs, rather than to doing the same with less. In a video interview this week, A Levie (CEO of Box, the popular cloud storage company) mentions the “lump of labor fallacy”, the idea that there is a finite amount of work, that machines can then “steal” from humans. History shows that technology adoptions don’t actually work that way, and Levie believes that AI will not be different. In the video he shares his views about how AI can be used to accelerate production processes (e.g. doing in 1h what has traditionally required 1-2 weeks), and about how this will lead to new / more work for current employees
The expected “devastating” effect of AI on radiologists’ jobs has not happened (yet). This week the New York Times published an analysis describing how radiology professionals still remain in high demand, when many experts were expecting them to be one of the first professions to be deeply affected by AI. Yes, there are plenty of AI systems that can automate radiologists’ jobs, but they are largely being used as a complement (“a second set of eyes”) to human doctors. Reading the article, there is not a very convincing reason to see this as a sustainable situation in the long term, but the point is that, even if there could finally be job substitution, it will take time…
Human writing may also be difficult to automate, at least in some contexts. This week famous author S Levy, now at Wired, published a summary of a recent speech he gave at Temple University, addressing Liberal Arts’ students. He also had an optimistic message with respect to AI, which he does not expect to be able to fully replace humans currently working in the humanities. To support his claims, he mentioned a 2023 research paper that showed how, In blind experiments, human beings valued what they read more when they thought it was from fellow humans and not a sophisticated system that fakes humanity
9 - Budget cuts at NASA confirm that leadership in space exploration is shifting to the private sector. Trump administration’s new federal budget proposal contemplates the cancellation of three major NASA programs, and analysts already see this as “the most radical remaking.in two decades” for the agency. Although the reductions still have to be approved by Congress, all this is perceived as a confirmation of an overall trend to shift the leadership in space programs to the private sector, with SpaceX as the key beneficiary
10 - Chips remain at the center of the tech Cold War, and it’s not clear that China is losing
A recent change in the export rules for chips will make it easier to sell to some countries, but not to China. Biden’s “AI Diffusion Rule”, which established three tiers for countries globally, and set different limitations for each of the tiers, is now being rescinded. It will be substituted by bilateral agreements with each country, which are seen as a more flexible mechanisms that will actually help tech companies increase their sales in many countries, including Saudi Arabia (this is linked to the country’s recent announcements on its AI plans). However, the change will not have any significant impact on sales to China
Recent news show how the pressure on China could even increase. According to the FT, Trump would be evaluating to add a new set of Chinese chipmakers to the technology export blacklist, including subsidiaries of giants like SMIC and YMTC, that are already in the list. At the same time, the US is looking to introduce new mechanisms to penalize countries that buy and use chips built by Huawei, in an effort to decelerate the company’s progress in its current trajectory to catch up with Nvidia
But it is not at all clear that China is losing
SMIC, a leading local chipmaker, is growing fast. The WSJ published a video this week arguing how the commercial / political pressures from the US “appear to have had the opposite effect”, and the company would now be contributing to “threaten US AI dominance”
Even Nvidia is trying to protect its business in China. They have just announced the deployment of a new R&D facility in Shanghai, with the purpose to adapt the company’s products (the ones they are allowed to sell in China) to the local market
LINKS:
1 - Population & natural resources
Biotech
New drugs
New developments in immunotherapy could have a significant impact on how we treat cancer. Immunotherapy has been used to attack tumors for more than 10 years, but now scientists talk about “a new era in cancer treatments” after some recent developments, which could be useful to address challenges in previous treatments. The innovations include the use of drugs with radioactive substances, drugs targeting genes that regulate tumor growth, and therapeutic vaccines. A new era in cancer therapies is at hand
Genomics
World’s first personalized Crispr (gene-editing) treatment saves the life of a baby. The ten-months old baby was suffering a mortal genetic disease, and he has apparently recovered after being administered three doses of a gene-editing treatment to mend the mutation that was causing the problem. World’s first personalized CRISPR therapy given to baby with genetic disease
Space
Budget cuts at NASA confirm that leadership in space exploration is shifting to the private sector. Trump administration’s new federal budget proposal contemplates the cancellation of three major NASA programs, and analysts already see this as “the most radical remaking.in two decades” for the agency. If Congress actually cancels the SLS rocket, what happens next?
SpaceX is expected to reaccelerate the development of its Starship rocket. They have apparently been able to solve issues first identified in an unsuccessful test flight back in January. There is a new test flight programmed for next week, and the plan is to get as close as possible to the initial target of 25 test flights during 2025. This is the rocket chosen by NASA for the coming mission to the Moon, and Elon Musk has also shared his ambition to use it for a mission to Mars. SpaceX Tests Starship Fixes After Back-to-Back Failures
Startups are proliferating in the space industry, including the building of innovative rockets. An example is Gilmour Space, an Australian company that is planning to test its Erin rocket soon, after aborting a test-flight last week due to an accident. The company was founded in 2012 and has raised $90m from VCs. The top fell off Australia’s first orbital-class rocket, delaying its launch
Energy
Data center energy demand could increase electricity prices for all. Grid upgrades driven by data center energy demands could make electricity more expensive for consumers, too, according to a report Wood MacKenzie, an energy research firm. The problem is that the money paid by large technology companies is expected not to be enough to pay for the upgrades, so there will be a penalty either for consumers or for the shareholders of the utilities. Data Centers’ Hunger for Energy Could Raise All Electric Bills
But there are also ways to exploit data centers for consumers benefit… A Bloomberg article describes how the connection of computing facilities with district heating systems can make it possible to exploit heat from the computers to warm homes. This is currently being tested in Sweden and Finland. Power-Hungry Data Centers Are Warming Homes in the Nordics
Nuclear
South Korea emerges as a global leader in nuclear technology:
South Korea is in a good position to benefit from the global nuclear energy revival under way. South Korea has the best large-scale atomic-energy industry outside China and Russia. Amid current plans across many countries to triple global nuclear energy production for 2050, under pressure from AI labs, South Korean companies expect to export 10 reactors by the end of this decade, including a recently delivered $20bn project in the Emirates. As Nuclear Power Makes a Comeback, South Korea Emerges a Winner
The next step will be innovative Small Modular Reactors (SMRs). The country is expecting to finalize the design of its first SMR this year, and to be able to deploy it by the mid-2030s. The vision behind SMRs is to reduce costs and times to deploy by standardizing designs. South Korea Goes Small in Bid for Nuclear Power Ascendance
Renewables
China is on its way to becoming the world’s first “electrostate”, largely driven by the government’s aspiration to become self-sufficient in energy, after a long period of dependence on foreign countries for oil and gas, which was perceived as a geo-strategic vulnerability. How Xi sparked China’s electricity revolution
2 - Efficiency & Productivity
New Transport Technologies
Electric Vehicles
Money keeps flowing to innovation projects for electric batteries:
CATL will have a (large) Hong Kong IPO to fund its global expansion. CATL, the world’s leading producer of batteries for electric cars, is planning its Hong Kong IPO, the largest so far this year, in which they expect to raise more than $4bn, that will help the company fund its global expansion. World’s largest EV battery maker CATL to raise at least $4bn
A new breakthrough by GM shows that battery technology is evolving fast. We have already reported here about massive improvements by Chinese vendors in charging times and ranges. This week we learned about a General Motors’ innovation, with a new type of low-cost battery for electric trucks. GM unveils new EV battery tech that could offer more than 400 miles of range and lower costs
Autonomous Cars
Chinese robotaxis are coming to Europe. Last week we discussed the plans of WeRide, a Chinese robotaxi startup, which is partnering with Uber for global expansion, including Europe. This week, the WSJ shares the plans of Baidu, the Chinese tech giant, to also deploy robotaxis in Europe (starting by Switzerland). Baidu Eyes European Debut for Driverless Taxi
Computing Infrastructure
Data Centers
Positive signs for AI infrastructure demand:
OpenAI makes Coreweave great again. Coreweave has just announced a large contract to provide “AI cloud” services to OpenAI. The deal has an estimated value of $4bn, to be paid in cash during the next 4 years. Coreweave’s stock has reacted positively. OpenAI Expands CoreWeave Tie-Up With New $4 Billion Cloud Deal
The deal comes to the rescue of Coreweave, which was under investor pressure, after missing expectations with its quarterly profit guidance. Investors are actually ultra-sensitive to anything related to the evolution of AI computing demand. CoreWeave shares fall as spending weighs on outlook
SoftBank, on its earnings call, confirmed its commitment to invest $100bn in the Stargate project to build AI data centers in the US, together with OpenAI and Oracle. This had previously been put into question because the company has not yet reached any kind of agreements with potential lenders and investors. SoftBank Stargate Venture With OpenAI Snags on Tariff Fears
India wants to be an AI infrastructure hub in SE Asia, with Airtel (the incumbent telecom operator) leading the effort. Airtel has just announced a plan to double the capacity of its data center business in less than 3 years, with an estimated investment of $600m. Low-cost India seen as potential regional hub in data centre boom
Communications
Starlink moves ahead to become a “default” global connectivity provider. This week we learned about the discussions they’re having with Emirates to provide the in-flight WiFi service. Emirates in Talks with Musk’s SpaceX to Get Starlink on Flights
Chips
Nvidia’s stock rises as concerns on demand for GPUs decrease. The share price is on track to its best month in a year (+44% from an April low). Among the catalysts, the recent changes in US chip export rules (opening the door to a huge deal in Saudi Arabia) and Big Tech CapEx commitments. Nvidia Shares Roar Back as Clouds Hanging Over Chipmaker Fade
Chinese players, like SMIC, are catching up… The WSJ published a video this week arguing how SMIC, the largest Chinese chip company, would now be contributing to “threaten US AI dominance”, after the commercial / political pressures from the US “appear to have had the opposite effect”. How China’s Biggest Chipmaker, SMIC, Could Threaten U.S. AI Dominance
Neuromorphic Computing
Researchers at RMIT University (Australia) develop a “neuromorphic” vision system, able to detect hand movement, store memories and process information like a system of human neurons, without the need for an external computer. This product could reduce the amount of energy needed to perform complex visual tasks compared with digital technologies used today. Tiny device promises new tech with a human touch
Artificial Intelligence
Microsoft: lots of uncertainties after its “divorce” from OpenAI. A long Bloomberg article this week discusses how Microsoft still has to define its role in the emerging AI landscape, after his progressive separation (still under way) from OpenAI. This is more pressure for S Nadella, specially with investors now asking where are the returns of the company’s previous investment ($13.75bn) in OpenAI. Microsoft’s CEO on How AI Will Remake Every Company, Including His
A key open issue: what stake will Microsoft own in the new OpenAI? This is currently being negotiated between the two companies, as OpenAI evolves from the previous structure (in which profits for investors were capped) to the new, more standard one. As we’ve said before, the negotiation is about economic rights, because the Non-Profit board will control the company in any case. OpenAI negotiates with Microsoft to unlock new funding and future IPO
AI: Apps, Agents
Google has announced the launch of a corporate VC unit to invest in AI startups. The AI Futures Fund seeks to invest in startups that are building with the latest AI tools from Google DeepMind. Google launches new initiative to back startups building AI
Agents
The race continues for AI labs to launch increasingly powerful “coding AI agents”
OpenAI is launching a new agent, “Codex”, specialized in generating code from natural language. The company is positioning the product as “a virtual co-worker for engineers”. ChatGPT is getting an AI coding agent
Also Google’s DeepMind presented AlphaEvolve, a new agent that aims to create code to tackle big problems in math and science, and which incorporates an “automatic evaluation” system to reduce hallucinations and increase accuracy. Google DeepMind creates super-advanced AI that can invent new algorithms
A team at the FT compared the performance of different AI agents on routine tasks. They tested products from OpenAI, Anthropic, Google and Meta. The overall conclusion is that fully autonomous agents are not here yet, but that current products already give an idea of how powerful these tools will be, specially for repetitive tasks. How AI agents compare on routine work tasks
B2C
Apple is exploring the use of AI to improve the iPhone’s battery life. They are working on an AI-powered battery management mode for iOS 19, which will analyze user behavior and make adjustments to conserve energy. Apple to Lean on AI Tool to Help iPhone Battery Lifespan for Devices in iOS 19
B2B
Meta’s AI lab (FAIR) has presented new prototypes using AI to facilitate scientific work. The PR highlights the company’s commitment to “achieving advanced machine intelligence (AMI) through focused scientific and academic progress”. Sharing new breakthroughs and artifacts supporting molecular property prediction, language processing, and neuroscience
AI being used in Brazil to spot new promising football players. There are new apps growing fast which young players use to upload videos showing them play, that then are screened by AI engines to identify the most promising ones. AI could be the Holy Grail in search for the next Neymar
Physical AI: Robots
Helsing, a German startup, is building unmanned mini-submarines for military surveillance. This is one more sign of the accelerating trend to use robots and AI for military applications. Lots of ethical questions coming, because it is clear that use cases won’t stop at surveillance. German defence start-up plans underwater drones for naval surveillance
AI: Foundational Models
Meta’s delays in its flagship LLM increase concerns that AI scaling laws may be hitting a wall. This week Meta announced a delay in the rollout of their flagship AI model, “Behemoth”. The release was initially expected for April 2025, then shifted to June, and now it has been pushed to the fall or later. This is seen as a sign that the approach to make AI progress through larger models (and data sets) may be running out of steam. Exclusive | Meta Is Delaying the Rollout of Its Flagship AI Model
LLMs perform better if complete instructions are provided at the start, rather than through a step-by-step conversation. This is the conclusion of a research paper by teams at Microsoft and Salesforce, that was shared this week. LLMs Get Lost In Multi-Turn Conversation
A friendlier explanation of what the paper says can be found in this twitter thread: elvis (@omarsar0) on X
Are Foundational Models turning into commodities? Competition is heating up. This week we learned that Harvey, a very successful AI app for lawyers, which was previously using OpenAI exclusively, will now also work with Anthropic and Gemini. Anthropic, Google score win by nabbing OpenAI-backed Harvey as a user
A key are of focus these days is increasing chatbots’s memory capacity. This aims to help models produce more accurate responses. But of course it can also help them better personalize conversations, and become more “sticky”, something to monitor, given recent claims about how addictive these apps have already become. Why memories are crucial for AI chatbots
Sakana, a startup, has released a new AI model architecture, which would be better for reasoning about unfamiliar problems. Sakana introduces new AI architecture, ‘Continuous Thought Machines’ to make models reason with less guidance — like human brains
AI: Security & Safety
Can AI agents trigger an “intelligence explosion”? There is no consensus about this. A new paper published this week analyzes all the recent algorithmic improvements in AI and reaches the conclusion that access to massive amounts of computing would be key to make disruptive progress in AI. As a consequence, controlling or restricting access to computing resources could be a way to reduce safety risks. How Fast Can Algorithms Advance Capabilities?
As expected, AI-generated content is making ID verification increasingly important. This Wired article describes a future in which everyone will feel the need to verify every interaction they have online. Deepfakes, Scams, and the Age of Paranoia
Intelligence Augmentation
Brain-Computer Interfaces
Apple planning to make it possible to control iPhones through brain-computer interfaces. They are working with Synchron, a startup that competes with Neuralink, to build solutions aimed at people with disabilities. Exclusive | Apple to Support Brain-Implant Control of Its Devices
3 - Economic / Business trends
Tech & Geopolitics
Saudi Arabia wants to be a global leader in AI infrastructure:
Saudi Arabia has launched Humain, a new AI company, focused on developing AI technologies and infrastructure (i.e. data centers), using the most advanced computing equipment (i.e. Nvidia GPUs). Saudi Arabia launches AI venture Humain ahead of Donald Trump visit
Does the country have the right talent pool to deliver? Parmy Olson at Bloomberg questions this. She warns about a risk to end up with "unused server farms”, unless the Saudis manage to attract the a mix of entrepreneurs and developers to build actual products on top of all this infrastructure. The Saudis' Bet on AI Plumbing Needs More Plumbers
The announcement was made during a visit by Trump to the Gulf, and was simultaneous with news of massive deals for US hardware companies, like Nvidia and Cisco, which will be partners in the project. US tech firms secure AI deals as Trump tours Gulf states
Chips remain at the center of the tech Cold War:
A recent change in the export rules for chips will make it easier to sell to some countries, but not to China. Biden’s “AI Diffusion Rule”, which established three tiers for countries globally, and set different limitations for each of the tiers, is now being rescinded. U.S. Scraps ‘AI Diffusion’ Rule in Revamp of Biden-Era Chip Curbs
Recent news show how the pressure on China could even increase. According to the FT, Trump would be evaluating to add a new set of Chinese chipmakers to the technology export blacklist, including subsidiaries of giants like SMIC and YMTC, that are already in the list. Trump administration considers adding Chinese chipmakers to export blacklist
Huawei is a key target. At the same time, the US is looking to introduce new mechanisms to penalize countries that buy and use chips built by Huawei, in an effort to decelerate the company’s progress in its current trajectory to catch up with Nvidia. US warns against using Huawei chips ‘anywhere in the world’
Meanwhile, Nvidia is trying to protect its business in China. They have just announced the deployment of a new R&D facility in Shanghai, with the purpose to adapt the company’s products (the ones they are allowed to sell in China) to the local market. Nvidia plans Shanghai research centre in new commitment to China
Emerging economic trends: Jobs
Are jobs more resilient (to AI) than we initially expected?
According to Box’s CEO, AI will be used to do more with the same jobs, rather than to doing the same with less. He describes how AI can be used to accelerate production processes (e.g. doing in 1h what has traditionally required 1-2 weeks), which will lead to new / more work for current employees. 🔮 How to build an AI-first company
The expected “devastating” effect of AI on radiologists’ jobs has not happened (yet). Yes, there are plenty of AI systems that can automate radiologists’ jobs, but they are largely being used as a complement (“a second set of eyes”) to human doctors. A.I. Was Coming for Radiologists’ Jobs. So Far, They’re Just More Efficient.
Human writing may also be difficult to automate, at least in some contexts. This week famous author S Levy, now at Wired, published a summary of a recent speech he gave at Temple University, in which he argued that human writing will remain differential and valuable. To support his claims, he mentioned a 2023 research paper that showed how, In blind experiments, human beings valued what they read more when they thought it was from fellow humans and not an AI. No, graduates: AI hasn't ended your career before it starts