one of the best of 2024, and the remainder | EUROtoday
Get real time updates directly on you device, subscribe now.
2024, yr of synthetic intelligence? While Open AI has simply introduced a man-made intelligence mannequin referred to as “o3” and able to taking the time to suppose towards itself, which suggests this nonetheless hypothetical second and nicknamed synthetic basic intelligence (AGI, in English ) throughout which the machine surpasses man in virtually all of its cognitive capabilities, the final twelve months have been wealthy in occasions. Selection of probably the most notable.
First right here is 5 excellent news:
- Accelerated drug and supplies design
In 2024, the AlphaFold software program, developed by DeepMind, has made nice progress in biochemistry by permitting the speedy prediction of the three-dimensional construction of proteins from their amino acid sequence. This advance, which guarantees to speed up the seek for new medicine, in addition to new supplies, was the topic of a publication on May 8 within the journal Nature.
The module for visualizing the construction of proteins, DNA, RNA, ligands and different molecules, and their interactions, begins with a cloud of atoms and steadily converges in the direction of probably the most seen molecular construction. exact. The creators of AlphaFold, Demis Hassabis and John Jumper, had been additionally awarded the Nobel Prize in Chemistry in 2024.
- Prediction of pure disasters
Models, reminiscent of Google’s Flood Hub, have been deployed in 80 nations to identify and report flood dangers. These programs analyze local weather information in actual time, enabling early evacuations and implementing preventative measures, thereby saving lives and lowering property harm.
The Chinese Huawei can also be engaged on this kind of prediction to anticipate local weather disasters. Problem: when there’s not sufficient information, as was just lately the case in Mayotte, this kind of mannequin loses its effectiveness.
- Creating reasonable movies
From Runway AI to Stability AI by way of Midjourney and even the Chinese Kling AI: increasingly software program makes it doable, in change for a paid subscription, to provide reasonable movies in a couple of minutes, which may shake up manufacturing movies and the creation of promoting.
Better nonetheless, offered you’re employed with good historians, such instruments make it doable to reconstruct, in a particularly visible approach, little-known episodes of the previous. This is what American director Dave Clark risked, who succeeded in bringing the story of the 320 to lifee barrage balloon battalion, a staff of American troopers, who labored, on June 6, 1944, to stop assaults by German plane at low altitude, on the D-Day seashores.
- Agriculture optimized by AI
Artificial intelligence is more and more built-in into agricultural programs. For instance, AI-equipped drones monitor crop progress, establish water and fertilizer wants, and detect illnesses at an early stage.
AI additionally supplies entry to new mapping of the territory as we see via this very good program referred to as Lidar HD. What can we see there? Driven by the IGN, this program goals to supply 3D information regarding the bottom and subsoil which might be homogeneous, wealthy, dependable and accessible to all.
- Artificial intelligence takes the time to suppose
This is a development began by the brand new mannequin referred to as “o3” introduced by OpenAI. Although it is going to solely be accessible to everybody from January, this AI instrument has taken the step of “thinking against itself” earlier than giving its end result. The consumer will be capable of select between three ranges of reflection: low, medium or excessive. The greater the extent of reflection, the longer the mannequin might take to formulate a response. Granted, this may end up in latency of some seconds. But, based on the New Scientisto3 additionally achieved a report rating of 75.7% on the Abstraction and Reasoning Corpus, a acknowledged AI reasoning take a look at, however has not but met the necessities of the “Grand Prix” which requires accuracy of 85%. Without the computational price necessities imposed by the take a look at, the mannequin additionally achieves a brand new report of 87.5%, when, on common, we people rating 84%.
And by way of threats, what ought to we be careful for?
This acceleration of progress in AI comes with extra worrying information. Among the threats of 2026 are:
- Failure to respect mental property
Generative AI, reminiscent of deepfake fashions, typically exploit copyrighted works with out the consent of the authors. For instance, manipulated movies of public figures have circulated, damaging their picture and elevating authorized questions on copyright safety. As a end result, processes have multiplied in current months towards main gamers within the sector from Open AI to Perplexity.
- Misinformation measurement XXL
Deepfakes have reached an unprecedented degree of realism, making it simpler to unfold false info. Sometimes it is for one of the best. So the movie Indiana Jones or the dial of future contains a 40-year-rejuvenated Harrison Ford in a struggle scene on a practice.
But whereas one in two Earthlings was referred to as to the polls in 2024, pretend movies of candidates had been shared massively this yr. “Today, disinformation spreads at high speed on the Internet, because false content, generally more attractive than true content, is encouraged by the economic models of large digital platforms which rely on the attention of Internet users,” explains the The Academy of Technologies which has simply revealed an inventory of suggestions to strengthen our crucial considering within the face of pretend information. Among these, that of “adding, in the Defense Code, a system of sanctions for disinformation operations harming the fundamental interests of the nation and orchestrated by foreign powers.”
- The menace of job loss accelerated by automation
AI has automated duties in numerous industries, resulting in job losses. Same worry amongst taxi and truck drivers. In February 2024 in San Francisco, protesters attacked a robotic taxi. A lot of professions are affected.
In the dubbing business, for instance, AI-generated artificial voices are changing actors, threatening 1000’s of jobs. And, in June, the French Society of Translators (SFT) printed a place paper warning concerning the penalties of generative AI on the interpretation professions.
- leveraging AI for surveillance functions
There is a powerful temptation for authoritarian regimes to make use of AI to raised monitor and repress opponents. Facial recognition programs establish protesters, resulting in arbitrary arrests.
Certainly, these measures usually are not the identical in all nations, however, in 2024, China has continued to enhance its surveillance programs based mostly on AI, specifically by growing applied sciences able to predicting, from gigantic databases, the behaviors of residents, like to notice them.
- In this loopy race, will the brand new gamers nonetheless have their say?
To Discover
Kangaroo of the day
Answer
The want for computing energy reinforces the specter of seeing the ability of huge expertise firms concentrated, limiting entry to innovation for smaller gamers. From Microsoft, which had the intelligence to put money into Open AI, to Amazon, which wager on Anthropic, Claude’s mother or father firm, by way of Google: cloud gamers have likelihood of rising as massive winners from this loopy race.
Of course, from Baidu to Tencent, not forgetting Kai-Fu Lee who launched 01.AI, China won’t stay idle in 2025 both. And the shock may come from Africa with magnificent initiatives, like Lelapa. It’s as much as European gamers from Mistral AI to Aleph Alpha by way of the Kyutai laboratory based mostly in Paris, to indicate that the Old Continent additionally has its say.
https://www.lepoint.fr/high-tech-internet/intelligence-artificielle-en-2024-nos-tops-et-nos-flops-26-12-2024-2578691_47.php