Artificial intelligence continues to dominate the software world this year. But more than ever, people are struggling with both its amazing capabilities and its amazing shortcomings. A new study found that AI agents doubling the duration of the tasks they can complete every seven months—an astonishing rate of exponential growth. But the quality of their work still suffers: the success rate for the most difficult tasks is about 50 percent. Chatbots help programmers and even write code themselves, but this may not help solve the biggest and most expensive problems. IT failureswhich arise from management errors that have remained constant over the past twenty years or more.
The energy needs of AI remain a major concern. To try to alleviate the situation, the startup is working to reduce the amount of heat generated during calculations by making the calculations reversible. Another builds a real computer human brain cells that can perform tests on drug candidates. And some are even considering sending data centers To moon.
iStock
Although the rankings of programming languages this year were quite predictable – yes, Python still number one – the future software development as vague as possible. As AI chatbots help many solve coding problems or simply code themselves, collecting reliable data about exactly what's going on becomes increasingly different. software engineers We work every day. People no longer post their questions on StackExchange or a similar site – they simply ask the chatbot.
This year list of the best programming languages does its best to work with this limited data, but it also raises the question: in a world where AI writes most of our code, how will programming Do languages change? Will we even need them or will AI just create optimized assembly code without the need for abstraction?
Eddie Guy
Robert Charettelifelong technologist and frequent IEEE spectrum participant, wrote back in 2005 all the known and preventable reasons why software development projects end in disaster. Twenty years later, nothing has changed, except for the additional trillions of dollars lost due to software failures. In this screed of more than 3500 wordsCharette lists numerous case studies, backed by statistics, highlighting the dismal state of IT governance as it is practiced—and still is—today. And to top it all off, he explains why AI won't come to the rescue.
Cortical laboratories
Australian startup Cortical Labs announces sale a biocomputer running on 800,000 living human neurons on a silicon chip. For $35,000, you get what amounts to a mini-brain in a box that can learn, adapt, and respond to stimuli in real time. The company has already proven its concept by teaching lab-grown brain cells to play. Pong (they are often superior to standard AI algorithms effectiveness of training). But the real application drug discovery. This “little brain in a vat,” as one scientist puts it, allows researchers to test whether experimental drugs restore the function of damaged neuronal cultures.
Model evaluation and threat research
It is difficult to agree on a consistent way of assessing how well large language models (LLM) work. Non-profit research organization Model evaluation and threat research (METR) proposed an intuitive metric—tracking how long it would take a human to complete tasks that AI could do. According to this indicator LLM opportunities double every seven months. If this trend continues, by 2030 the most advanced models will be able to quickly cope with tasks that currently take people a whole month of work. But for now, AI isn't always good at its job—the chance of getting the job done right for the longest, most complex tasks is about 50 percent. So the question is: How useful is a fast, cheap employee who produces garbage about half the time?
Edmond de Haro
There is a surprising principle that ties all software to the basic physics of hardware: erasing small amounts of information on a computer necessarily requires an expenditure of energy, usually lost as heat. The only way to avoid losing this energy is to never erase information. This is the basic idea behind reversible computing, an approach that remained in the academic field until this year.
After three decades of academic research, reversible computing will finally become commercial with startup Vaire Computing. Wire's first prototype chip recovers energy in an arithmetic circuit. The team claims that with their approach they could eventually achieve a 4,000x increase. energy efficiency an improvement over conventional chips. The catch is that this requires new gateway architectures, new design tools and integrations MEMS resonators on the chip. But a prototype is already in development. reversible computation went from “interesting theory” to “we’re actually building it.”
Nicole Millman
Apache air flowan open-source workflow orchestration software originally created by Airbnb, was all but dead by 2019. Then an open source enthusiast came across it while working at Internet of Things and thought, “This is too good to die.” He brought the community togetherand by the end of 2020 they released Airflow 2.0. Now the project is thriving. It boasts 35 to 40 million downloads per month and over 3,000 members worldwide. And Airflow 3.0 launched with a modular architecture that can run anywhere.
iStock/IEEE Spectrum
In 2004 President Bush set a goal For United States go to electronic medical records (HER) by 2014, promisingly transformed healthcare and huge cost savings. Twenty years and more than $100 billion later, we have achieved widespread EHR adoption.and created another nightmare. Doctors now spend an average of 4.5 hours a day staring at screens instead of patients, and navigating poorly designed software systems.
Rushing to implement EHRs before they were ready meant ignoring warnings about systems engineeringcompatibility and cybersecurity. We are now faced with fragmented systems that don't communicate with each other (the average hospital uses 10 different EHR providers internally) and physicians are experiencing record levels of burnout. And to top it all off, data leak since 2009, 520 million records have been exposed. Health care spending did not fall as promised: it reached $4.8 trillion, or 17.6 percent of GDP. Irony? Artificial intelligence scribes are now being developed to solve the problems created by the latest generation of technology, allowing doctors to look at patients instead of their keyboards again.
Intuitive machines
Whether space or lunar data centers are a promising trend or a fever dream is the subject of much debate. However, earlier this year Lonestar Data Holdings sent a mini data center weighing 1 kilogram and 8 terabytes to the Moon aboard the Intuitive Machines lander. The goal is to protect confidential data from earthly disasters (submarine cable cuts, hurricaneswar) and take advantage of a loophole in data sovereignty laws – since the Moon is not under the jurisdiction of any country, you can post black boxes according to the laws of any country you want.The lunar surface has permanently shadowed craters with a temperature of -173°C, which may facilitate cooling (although the lack of an atmosphere makes thermal radiation experiencing). Nearby sun-drenched peaks will provide solar energy. Governments are interested: Florida and the Isle of Man already store data there. But the problems are obvious: the 1.4 second latency rules out real-time applications, it takes a mission to the moon to fix anything, and the throughput is terrible.
Articles from your site
Related articles on the Internet
