A discussion of recent news and events.
Viral Studio Ghibli-style AI images showcase power – and copyright concerns – of ChatGPT update | CNN
Read original at CNN →<DIV><section data-editable="main" data-track-zone="main" data-reorderable="main"> <article data-uri="cms.cnn.com/_components/article/instances/cm8s5hizb004q2cp2dxfj3tce@published" role="main" data-drag-disable="true" data-unselectable="true" data-regwall-disabled="false" data-subscription-only="false" data-paywall-disabled="false"><section data-tabcontent="Content"><main><div data-editable="content" itemprop="articleBody" data-reorderable="content"><p><cite><span data-editable="location"></span><span data-editable="source">CNN</span> — </cite></p><p data-uri="cms.
cnn.com/_components/paragraph/instances/cm8s5hizb004p2cp22ax2culj@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">Just days after OpenAI launched its most advanced AI image generator to date, a social media trend imitating the work of Japanese animation company Studio Ghibli is demonstrating both the technology’s power and the copyright concerns it raises.
</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00013b5vauido25o@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">The latest update to GPT-4o, released Tuesday, features many practical advancements, including more accurate text rendering and the ability to follow more detailed, complex prompts.
But it has also been trained at length on a “vast variety of image styles,” according to a post on OpenAI’s website, stunning users with its ability to generate still images and videos reminiscent of their favorite animations, from “South Park” to classic <a href="https://x.com/BennettWaisbren/status/1905247775190864381" target="_blank">claymation.
</a></p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00023b5v0i7mwj0j@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">But one style quickly flooded X and Instagram, as users of ChatGPT (and OpenAI’s text-to-video service, Sora) began emulating the work of beloved animation studio behind movies like “Spirited Away” and “Howl’s Moving Castle.
”</p><div data-editable="settings" data-url="https://media.cnn.com/api/v1/images/stellar/prod/mmdspaw-ec001.jpg?c=original" data-original-width="2790" data-original-height="1508" data-original-ratio="0.5405017921146953" data-observe-resizes="" data-component-name="image" data-name="MMDSPAW_EC001.jpg" data-uri="cms.
cnn.com/_components/image/instances/cm8s83htw00023b5vqtat4k18@published" data-image-variation="image" data-breakpoints="{"image--eq-extra-small": 115, "image--eq-small": 300, "image--eq-large": 660}"><picture><source height="1508" width="2790" media="(max-width: 479px)" srcset="https://media.
cnn.com/api/v1/images/stellar/prod/mmdspaw-ec001.jpg?q=w_680,c_fill/f_webp" type="image/webp"><source height="1508" width="2790" media="(min-width: 480px)" srcset="https://media.cnn.com/api/v1/images/stellar/prod/mmdspaw-ec001.jpg?q=w_1160,c_fill/f_webp" type="image/webp"><source height="1508" width="2790" media="(min-width: 960px)" srcset="https://media.
cnn.com/api/v1/images/stellar/prod/mmdspaw-ec001.jpg?q=w_1015,c_fill/f_webp" type="image/webp"><source height="1508" width="2790" media="(min-width: 1280px)" srcset="https://media.cnn.com/api/v1/images/stellar/prod/mmdspaw-ec001.jpg?q=w_1110,c_fill/f_webp" type="image/webp"><img src="https://media.cnn.
com/api/v1/images/stellar/prod/mmdspaw-ec001.jpg?q=w_1110,c_fill" alt="A still from 2001's "Spirited Away" of Haku (in dragon form) and Chihiro by Studio Ghibli." onload="this.classList.remove('image__dam-img--loading')" onerror="imageLoadError(this)" height="1508" width="2790" loading="lazy"></picture></div><p data-uri="cms.
cnn.com/_components/paragraph/instances/cm8s62yvy00033b5vkx1qh4f2@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">Some recreated scenes from pop culture or politics in the Japanese company’s iconic style, including a <a href="https://x.com/PJaccetturo/status/1905151190872309907" target="_blank">reworked trailer</a> for “The Lord of the Rings: The Fellowship of the Ring,” <a href="https://x.
com/timeimmemorial_/status/1905262678521582027" target="_blank">scenes</a> from “The Sopranos,” and Donald Trump and JD Vance’s heated real-life White House exchange with Ukrainian President Volodymyr Zelensky.</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00043b5v3uathqhp@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">Unsurprisingly, some of the most viral posts put a Ghibli spin on popular memes, including the “<a href="https://x.
com/heyBarsee/status/1904891940522647662" target="_blank">distracted boyfriend</a>,” the “bro explaining” meme (pictured top) and the infamous image of <a href="https://x.com/venturetwins/status/1904915503505670246" target="_blank">Ben Affleck smoking</a>. Another viral X post depicted the platform’s owner, Elon Musk, <a href="https://x.
com/Jason/status/1905031350681321867" target="_blank">playing with cutlery</a> — an image based on the recent video of the billionaire balancing spoons during a dinner hosted by Trump in New Jersey.</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00053b5vbtebwseu@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">Also widely shared, however, is a 2016 <a href="https://www.
youtube.com/watch?v=ngZ0K3lWKRc" target="_blank">video</a> in which Studio Ghibli co-founder Hayao Miyazaki describes AI-generated art as an “insult to life itself.” Miyazaki is known for his hand-drawn animation and painstaking frame-by-frame method.</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00063b5v6yitjofr@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">“I am utterly disgusted,” he says in the video, responding to a video of a monster character generated using text prompts.
“If you really want to make creepy stuff, you can go ahead and do it, but I would never wish to incorporate this technology into my work at all.”</p><div data-editable="settings" data-url="https://media.cnn.com/api/v1/images/stellar/prod/2014-11-09t120000z-1478189358-gm1eab91hqd01-rtrmadp-3-film-governorsaward.
jpg?c=original" data-original-width="2096" data-original-height="1600" data-original-ratio="0.7633587786259542" data-observe-resizes="" data-component-name="image" data-name="2014-11-09T120000Z_1478189358_GM1EAB91HQD01_RTRMADP_3_FILM-GOVERNORSAWARD.jpg" data-uri="cms.cnn.com/_components/image/instances/cm8s87wip00043b5vljie78o5@published" data-image-variation="image" data-breakpoints="{"image--eq-extra-small": 115, "image--eq-small": 300, "image--eq-large": 660}"><picture><source height="1600" width="2096" media="(max-width: 479px)" srcset="https://media.
cnn.com/api/v1/images/stellar/prod/2014-11-09t120000z-1478189358-gm1eab91hqd01-rtrmadp-3-film-governorsaward.jpg?q=w_680,c_fill/f_webp" type="image/webp"><source height="1600" width="2096" media="(min-width: 480px)" srcset="https://media.cnn.com/api/v1/images/stellar/prod/2014-11-09t120000z-1478189358-gm1eab91hqd01-rtrmadp-3-film-governorsaward.
jpg?q=w_1160,c_fill/f_webp" type="image/webp"><source height="1600" width="2096" media="(min-width: 960px)" srcset="https://media.cnn.com/api/v1/images/stellar/prod/2014-11-09t120000z-1478189358-gm1eab91hqd01-rtrmadp-3-film-governorsaward.jpg?q=w_1015,c_fill/f_webp" type="image/webp"><source height="1600" width="2096" media="(min-width: 1280px)" srcset="https://media.
cnn.com/api/v1/images/stellar/prod/2014-11-09t120000z-1478189358-gm1eab91hqd01-rtrmadp-3-film-governorsaward.jpg?q=w_1110,c_fill/f_webp" type="image/webp"><img src="https://media.cnn.com/api/v1/images/stellar/prod/2014-11-09t120000z-1478189358-gm1eab91hqd01-rtrmadp-3-film-governorsaward.jpg?q=w_1110,c_fill" alt="Japanese film director and animator Hayao Miyazaki poses during the Academy of Motion Picture Arts and Sciences Governors Awards in Los Angeles, California, in 2014."
onload="this.classList.remove('image__dam-img--loading')" onerror="imageLoadError(this)" height="1600" width="2096" loading="lazy"></picture></div><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00073b5v1ba6gk48@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">OpenAI’s updated image generator has also prompted renewed discussions over the role of AI and art.
It comes just weeks after nearly 4,000 people <a href="https://www.cnn.com/2025/02/11/style/christies-ai-auction-open-letter-tan/index.html">signed an open letter</a> calling on Christie’s auction house to cancel a first-of-its-kind sale dedicated solely to AI art over concerns that the programs used to create some generative digital pieces are trained on copyrighted work and exploit human artists.
</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00083b5vrrqh0b0f@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">OpenAI’s CEO Sam Altman made light of the trend on X, <a href="https://x.com/sama/status/1904921537884676398" target="_blank">joking</a> that after “a decade trying to help make superintelligence to cure cancer or whatever” it was Studio Ghibli images that had generated viral interest in his work.
</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy00093b5v50l0j2mv@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">“Mostly no one cares for first 7.5 years, then for 2.5 years everyone hates you for everything,” he wrote. “Wake up one day to hundreds of messages: ‘Look I made you into a twink Ghibli style haha’” Altman added, referring to a gay slang term for men who are young, boyish and slim.
</p><p data-uri="cms.cnn.com/_components/paragraph/instances/cm8s62yvy000a3b5vxfvc1mcb@published" data-editable="text" data-component-name="paragraph" data-article-gutter="true">As is often the case with AI-generated art, the images raise various copyright questions — not only around Studio Ghibli’s work but of the images being reimagined.
When CNN prompted ChatGPT to reproduce some of the Ghibli-style memes, the service refused, saying that ” the request didn’t follow our content policy.”</p><p data-uri="cms.cnn.com/_components/footnote/instances/cm8s5j3uj00083b6mml77kjar@published" data-editable="text" data-article-gutter="true">CNN’s Jacqui Palumbo contributed to this story.
</p></div></main></section></article></section></DIV>
AI-Generated Art Gets Its Own Sleek Print Magazine
Read original at Forbes →The newly launched AI Art Magazine features artwork from around the world, including this image from ... [+] Irish conceptual artist Kevin Abosch.The AI Art MagazineArt made with artificial intelligence is ubiquitous online, appearing on platforms from Instagram and Reddit to websites hosting generative AI tools themselves.
Now, artists have a new and somewhat ironic outlet for showcasing their digital, algorithm-assisted creations: paper.A polished new print publication, The AI Art Magazine, focuses entirely on AI and art, as its name suggests. The first issue, which came out Dec. 6, spans 176 pages filled with artwork submitted in response to an international open call and selected by jurors — artists, graphic designers, technologists and one AI-generated judge named Xiaomi that looks part robot, part anime character.
The independently funded biannual magazine, which is published in Germany, costs 22 euros (roughly $22.95) and doesn’t include ads, at least not yet.“This project is fueled by pure passion and a shared vision to chart the course for the future of AI art,” reads a magazine mission statement. “It is a collective effort, driven by a community that dares to imagine, innovate and inspire.
”The cover of the magazine's inaugural issue features work by Japanese artist Emi Kusano, who shares ... [+] her approach to AI in an interview.The AI Art MagazineThe jurors include Boris Eldagsen, who turned down a prestigious photography award after revealing he’d generated the winning image with AI to provoke debate.
In brief blurbs, the judges share why they chose to feature particular works, while accompanying essays and interviews allow the creators to elaborate on their artistic process, what excites them about incorporating AI into their work and their views on its strengths, weaknesses and limitations.“It is a world-shaping technology that needs the voices of artists to engage with it critically and creatively, to ask where it might fail us and how it might better our lives,” artist Kevin Esherick says in a Q&A.
“The best way to understand these technologies is to work with them.”Esherick was just 20 when his older brother died, and it’s a loss he feels daily. The piece featured in the magazine, a hazy, dreamlike image of his late sibling, is part of his deeply personal series “I’m With You,” which imagines a world where his brother is still here to share in life’s moments.
The Brooklyn-based artist trained an AI model on pictures of his brother, then disrupted the image generation process, leaving only a cloudy imprint of his form. Esherick produced more than 100,000 images for the collection, then narrowed them down to 24, naming each after a song his brother loved or would have loved.
In “Especially in Michigan 2024,” seen in the magazine, it’s as if his deceased family member exists behind a veil, in another realm of consciousness.“For me, these pieces are reflections on grief and memory, absence and presence, joy and hope,” Esherick says. “They’re about possibility, what could’ve been, and what lives on.
”The magazine itself centers on what’s possible at a time when artists are grappling with what AI means for them, their futures and creativity on the whole. The inaugural issue includes 50 images in all, with the next issue scheduled for the summer.Essays and interviews allow a slect number of featured creators to elaborate on their artistic ...
[+] process and what excites them about incorporating AI.The AI Art MagazineIn “Multiple Sclerosis – Ataxia,” Sabine von Bassewitz tries to visually convey the experience of a relapse marked by spasticity and movement difficulties, not to mention emotional upheaval. The work depicts an artist sitting on the floor drawing with a disembodied hand while wearing a shoe on one arm.
“I find it very difficult to describe the symptoms verbally in a way that my listeners can understand,” the German photographer says in the magazine. “I often get the impression that even the neurologist treating me doesn’t fully understand even though she is very familiar with the subject. Midjourney, on the other hand, seems to understand me.
”‘AI Is A Reality And Will Not Disappear’In the past couple of years, AI has fundamentally transformed the art landscape as widely used generative tools like Midjourney, Stable Diffusion and OpenAI’s Dall-E enable anyone to create images simply by inputting text prompts. This rapidly evolving field has sparked passionate, and often divisive, debate among creatives.
Some embrace machine learning as a tool that can steer them in weird and wonderful directions, while others are angry their work is being stolen from the internet to train AI datasets without credit or compensation. Many also fear it will steal their livelihoods.“I fully understand traditional artists' concerns,” Mike Brauner, the magazine’s publisher and co-founder, said in an email interview.
“However, you need to recognize that AI is a reality and will not disappear. We are at the very beginning of something groundbreaking, and I can only encourage every artist to familiarize themselves with the new possibilities.”ForbesGoogle Challenges Artists To Defy AI Cliches, With Striking ResultsBy Leslie KatzBrauner, founder of Hamburg-based creative agency Polardots Studio, co-founded the magazine alongside illustrator Christoph Grünberger, author of the book Age of Data: Embracing Algorithms in Art & Design, which focuses on the aesthetics and creativity of a new generation of designers turning to algorithm-supported tools.
The pair chose to produce the magazine with open-thread binding to contrast digitally created art with traditional bookmaking, “resulting in a striking homage to the fusion of modern technology and classic artistry.”While The AI Art Magazine is currently print-only, Brauner revealed plans to launch a companion online database where artists can display their work.
“We received an overwhelming number of submissions from 40 countries for our first open call,” he said, “and we really want to give those a stage to present themselves.”"We are at the very beginning of something groundbreaking," says Mike Brauner, publisher and ... [+] co-founder of the new biannual magazine.
The AI Art Magazine
Apple Intelligence vs Google Gemini vs Galaxy AI: what are the differences?
Read original at PhoneArena →The past couple of years have been a wild ride in the tech world. Namely because AI became such a huge buzzword that everyone wants on that particular bandwagon. Google announced a few different AI projects — from Bard to DeepMind to Gemini (it’s all unified under the Gemini name now, thankfully), Samsung released its With all of these manufacturers and developers throwing AI features at us, most of which cover the same ground, it quickly becomes a confusing mess.
Which one does what, who has exclusive features, who does it best?Let’s take a step back, take a deep breath, and look at all the currently known features for Apple’s AI, Google’s Gemini, and Galaxy AI.We will be including the promised features as well — since Apple is still in the middle of rolling out the complete Apple Intelligence feature set.
We figured it’s better to throw them in as well, so we can have a clearer picture of which platform is going for what, even if it’s not quite there yet.Apple AI - No | Gemini - Yes | Galaxy AI - Yes Both the Google Pixel and Galaxy phones offer live translation on a call between a few of the bigger languages.
It's not flawless, but can be helpful in certain scenarios.Smart replyApple AI - Yes | Gemini - Yes | Galaxy AI - PartialThe language models can pretty accurately interpret what a message you have received is about, and then generate a more elaborate reply for you. Apple AI and Gemini do this in email and text, Galaxy AI currently only does if in the Messages app (texts).
Compose and rewrite textApple AI - Yes | Gemini - Yes | Galaxy AI - YesYou can have the AI arrange your text neatly, compose entire paragraphs, or change the tone and style of a piece of text. Apple’s AI can also call up the help of ChatGPT for longer, more elaborate, essay pieces. In any case, all three assistants here cover this feature.
Image-related AI features Generate imagesApple AI - Yes | Gemini - Yes | Galaxy AI - PartialApple has its own Image Playground, which generates animation-style images that generally aren’t… very diverse or usable. But, again, with ChatGPT baked into iOS 18, you can just ask it to generate more realistic images, or ones in different styles.
It’s worth noting that Apple also has the new “Genmoji”, which will specifically generate emoji-style images based on your image prompts. These are… met with lukewarm reception, as they kind of defeat the purpose of expressing yourself with emoji — the fun is in their limited numbers and trying to create different combinations and meanings.
Google’s Pixel Studio is also a bit better at making images out of thin air, but it’s still a bit limited — specifically, it will refuse to do human figures. Google also has the Emoji Lab, which will mix the characteristics of two emoji into one to create a sort of Franken-moji. These are a bit more fun than Genmoji.
Samsung’s AI can currently generate images from your own hand drawings, but doesn’t currently have the feature to make them from text prompts, nor emoji generators.Visual searchApple AI - Yes | Gemini - Yes | Galaxy AI - YesApple AI has the new Visual Intelligence feature — press and hold the Camera Control button and then you can use the viewfinder to either ask ChatGPT or Google what you are looking at.
Both Google and Samsung phones have Gemini with access to the camera, which works much like Visual Intelligence. There’s also Circle to Search, where you can perform an image search of anything you currently have on your screen.Magic eraserApple AI - Yes | Gemini - Yes | Galaxy AI - Yes This has been on Samsung and Google phones long before it was attributed to AI.
Magic Eraser is a “generative eraser” tool, meaning it will delete any object from a photo and try to generate the missing background, based on context around it. Apple now has the Clean Up tool in Photos, which does much the same. One can argue that Apple’s tool is a bit more “AI”, since it will immediately identify and suggest which objects you may wish to delete from a photo.
Auto crop out subject from a photoApple AI - Yes | Gemini - No | Galaxy AI - Yes Again, this has been around before the AI craze. Essentially, you can tap and hold on an object in a photo, and the software will “identify” its boundaries and allow it to “lift” it from the photo. Then, drag it to paste into another app, or generate a copy of that image with a transparent background.
It’s a pretty useful tool for making YouTube thumbnails, and has been available on iPhone and Galaxies for a couple of years now.Pixel exclusive: Photo Unblur, Reimagine, Add me, Magic EditorApple AI - no | Gemini - yes | Galaxy AI - noGoogle’s been working on image-related tricks for some years now.
The Photo Unblur feature will analyze and remove the blur from any picture in Photos. Even if you uploaded an old camera photo!Magic Editor allows you to rotate and change the positioning of a photo. If you end up moving outside the lines of the original photo, Magic Editor will try to generate more background, based on the image.
Reimagine can change entire subjects in a photo based on a text prompt. Like “Make the grass yellow” or “Make the sky cloudy”. Add me is a cool way to take a group picture without needing outside help. One person takes a photo of the group, then someone else steps in and directs the original photographer to take their spot in the frame.
The AI will stitch picture 2 with picture 1 to make the entire group look as if they were together at the same time.Some of these may come to other Android phones via a Photos update — Unblur is available, Magic Editor is in testing, and Reimagine may also be pushed at some point. But Add me, for example, is specifically a Pixel Camera feature.
General AI featuresContextual understanding and multi-layered commandsGoogle’s Gemini is the leader here. The conversational Gemini Live model is currently available, and in general, Gemini’s capabilities are comparable to ChatGPT’s. You can have Google’s assistant answer questions from the web, brainstorm or lead a conversation with you, or reach for limited information within apps and your account.
Apple’s Siri is supposedly getting a major overhaul this March, where it will also be able to follow conversations but also do complex, multi-layered tasks. Like pull information from one app to apply it to a query for another app. For the time being, we do have access to CharGPT, which is quite deep and elaborate, straight through Siri.
Bixby is currently lagging behind, but from what we know about One UI 7, which should be launching with the Galaxy S25, Samsung’s assistant should be getting a lot smarter. Apparently, it will be powered by Samsung’s own LLM model — Gauss — and Google’s Gemini. We’ll see how that pans out, but there are huge promises for a conversational style and multi-app operations.
Exclusive AI features right nowApple IntelligenceApple put a huge emphasis on re-doing how notifications are delivered with the help of AI. For one, messages that are more “important” will be automatically pushed to the top of your feed. Secondly, AI will be used to summarize long messages or chat strings, so you can get a better idea of what’s going on at a glance.
At least that’s the general idea. The execution is not quite there yet, with some summaries being pretty vague or funny.The same treatment has been applied to the Email app, where a new Priority tab will be filtered and summarized by AI.Google GeminiGoogle did a couple of unique moves with its AI implementation.
One is the new Screenshots app — any screenshots you ever take will be collected here, and AI image recognition will slap a few tags onto it. So, you can easily search for it later. Like “Phone number of plumbing service” or “Poster of a show with dates on it”.The other unique feature is in none other than the Weather app.
In the same spirit as summarizing a long article, Gemini will give you a quick recap of the weather forecast at the very top. Yeah, the level of usefulness here is arguable.Then, we have a slew of other smart Google features, which were released long before everything had “AI” on it, but definitely fall under the umbrella:Call screeningDirect my callHold for meClear CallingReserve with GoogleSamsung Galaxy AISamsung started off 2024 strong, with a slew of features in Galaxy AI.
However, by the end of the year, it’s definitely lacking in unique or distinct features. Of course, this should change in early 2025 — we’ll see if Samsung manages to cook up something new and different, which the competition does not yet have.
Will 2025 Be a “Technology Wake-Up Call” for Clinicians?
Read original at Psychology Today →Source: DALL-E / OpenAIThe year 2025 may well mark a pivotal moment in the evolution of artificial intelligence (AI) in medicine. A new prepress study evaluating OpenAI’s GPT-4 and o1-preview model demonstrates that AI is not only achieving impressive feats in clinical reasoning but is doing so without supplemental training on domain-specific data.
This achievement represents a significant leap in what general-purpose large language models (LLMs) can accomplish, fueled by innovations in reasoning frameworks such as chain-of-thought (CoT) processing.The findings are both promising and provocative. On one hand, the o1-preview model excels in tasks requiring complex diagnostic and management reasoning, rivaling human clinicians.
On the other, it reveals critical gaps in probabilistic reasoning and triage diagnosis, areas where human expertise remains paramount. This duality raises important questions about how AI will integrate into medical workflows and redefine the role of clinicians.There's a lot to unpack here, and I suggest reading the study carefully as I'm only touching on some of the key points, particularly the results with the o1-preview model.
A Tale of Strengths and WeaknessesThe study evaluated the o1-preview model across five experiments, including differential diagnosis generation, diagnostic reasoning, triage differential diagnosis, probabilistic reasoning, and management reasoning. The results were adjudicated by physician experts using validated psychometrics, providing a benchmark for comparison against human controls.
StrengthsDifferential diagnosis generation: The o1-preview model achieved an 88 percent accuracy rate, far surpassing the 35 percent accuracy demonstrated by human clinicians in the same task. Its output was consistently rated as more comprehensive and precise, particularly in rare and complex diagnostic scenarios, where the model’s CoT reasoning allowed it to identify conditions often overlooked by clinicians.
Diagnostic and management reasoning: The o1-preview model displayed significant advancements in diagnostic and management tasks. In 84 percent of cases, the model’s reasoning was rated as on par with or exceeding that of human experts, who achieved comparable accuracy in only 64 percent of cases. Physicians praised the model’s structured and logical approach, which mirrored the stepwise critical thinking employed by clinicians and synthesized data from diverse clinical inputs to produce actionable recommendations.
LimitationsProbabilistic reasoning: The model struggled with tasks requiring nuanced probabilistic reasoning—a cornerstone of medical decision-making. While the o1-preview model’s performance was consistent with prior LLMs, human clinicians continued to excel in this area, demonstrating greater adaptability in assigning likelihoods to competing diagnoses and dynamically balancing risks in uncertain situations.
Triage differential diagnosis: No improvements were observed in triage tasks that require prioritizing cases by severity. While human clinicians achieved a 70 percent accuracy rate in these high-pressure, dynamic scenarios, the model’s logical but rigid outputs fell short, lacking the adaptive nuance required for real-time decision-making in emergency or critical care settings.
The Role of Chain-of-Thought ReasoningA standout feature of the o1-preview model is its reliance on chain-of-thought (CoT) reasoning, a framework that enables the AI to generate intermediate steps in its reasoning process before arriving at a final answer. This process allows the model to explain its thought process, making its outputs more transparent and easier for clinicians to interpret.
By breaking down complex problems into smaller steps, CoT reasoning reduces the risk of logical errors, particularly in tasks requiring critical thinking. Moreover, this approach mimics the way clinicians address diagnostic challenges—systematically considering symptoms, test results, and medical history to form conclusions.
The use of CoT reasoning may be an important factor in the model’s success with diagnostic and management reasoning, even as it struggles with the more dynamic aspects of clinical practice, such as triage.The Remarkable Absence of Supplemental Clinical TrainingAnother striking aspect of the o1-preview model is that it was not trained on supplemental clinical data.
Unlike earlier AI systems fine-tuned on medical data sets, o1-preview achieved its performance using general-purpose training. This accomplishment suggests that broad, general training data combined with advanced reasoning frameworks can rival domain-specific training, reducing the need for costly and time-intensive fine-tuning processes.
The absence of supplemental training also eliminates concerns about patient privacy, biased data sets, and overfitting to specific scenarios. However, it means the model’s performance is limited to patterns present in its general training data, leaving gaps in areas requiring contextual nuance. This highlights both the promise and the current limitations of generalist AI systems in specialized domains like healthcare.
The o1-preview model’s performance highlights both the promise and the limitations of LLMs in medicine. For clinicians, this study serves as a wake-up call: AI is no longer a futuristic concept—it’s here, and it’s redefining what is possible in patient care.AI as a partner: Models like o1-preview are not replacing clinicians but augmenting their capabilities.
They excel at tasks like differential diagnosis generation and management planning, freeing up clinicians to focus on patient interaction and decision-making.Closing the gaps: While o1-preview shines in structured reasoning tasks, its struggles with probabilistic reasoning and triage emphasize the irreplaceable value of human expertise.
These gaps point to opportunities for future AI development.The need for new benchmarks: Current evaluation methods, such as multiple-choice question benchmarks, fail to capture the complexity of real-world clinical scenarios. Robust, scalable benchmarks and clinical trials are essential to understand AI’s true potential in healthcare.
Digital Health and "Another" Inflection Point?The o1-preview model may represent a turning point in the integration of AI into medicine. And as we've heard this claim many times, its ability to perform superhuman reasoning tasks without supplemental clinical training is important—as an achievement and a challenge.
As AI continues to evolve, clinicians must adapt to this new reality, embracing AI as a cognitive partner while maintaining the human expertise that defines the art of medicine.2025 doesn't just represent a wake-up call; it may be the beginning of a new era. The question is no longer whether AI will transform medicine, but how clinicians and AI will work together to shape the future of healthcare.
Sam Altman on the Future of A.I. and Society
Read original at The New York Times →Listen to our interview with Sam Altman.More episodes ofDealBook SummitListen and follow DealBook SummitApple Podcasts | Spotify | Amazon MusicDealBook Summit includes conversations with business and policy leaders at the heart of today’s major stories, recorded live at the annual DealBook Summit event in New York City.
The co-founder and chief executive of OpenAI discusses how ChatGPT has changed how the world sees A.I. and the prospects of artificial general intelligence — a machine that can do anything the human brain can do — in the years to come.ImageCredit...The New York TimesFollow DealBook’s reporting at https://nytimes.
com/dealbookHosted by Andrew Ross Sorkin, a columnist and editor of DealBook, a daily business and policy report from The New York Times, DealBook Summit features interviews with the leaders at the heart of today’s major stories, recorded live onstage at the annual DealBook Summit event in New York City.
The DealBook events team includes Julie Zann, Caroline Brunelle, Haley Duffy, Elaine Chen, Angela Austin, Hailey Hess, Dana Pruskowski, Matt Kaiser and Yen-Wei Liu.Special thanks to Sam Dolnick, Nina Lassam, Ravi Mattu, Beth Weinstein, Kate Carrington, Isabella Anderson and Jeffrey Miranda.Sarah Kessler is an editor for the DealBook newsletter and writes features on business and how workplaces are changing.
More about Sarah KesslerAdvertisementSKIP ADVERTISEMENT
