🤖 🧠 Thibault GEOUI 🧬 💊

𝗨𝗻𝗹𝗼𝗰𝗸𝗶𝗻𝗴 𝗔𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝗔𝗜 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗢𝗽𝗲𝗻𝗔𝗜’𝘀 𝗣𝗿𝗼𝗷𝗲𝗰𝘁 𝗦𝘁𝗿𝗮𝘄𝗯𝗲𝗿𝗿𝘆 Artificial Intelligence is evolving rapidly, and OpenAI’s latest breakthrough—Project Strawberry—marks a significant step towards more advanced AI reasoning capabilities. This project aims to enhance problem-solving by enabling AI to perform complex, multi-step tasks with a more human-like approach to reasoning. What’s New with Strawberry? Unlike its predecessors, Strawberry allows AI models to think more deeply about problems before offering solutions. Whether solving advanced algebra, planning marketing strategies, or answering complex physics questions, Strawberry models take more time to process inputs, mimicking deliberate, human reasoning. This new method represents a substantial leap from sequential generation approaches, enabling more thoughtful and accurate responses. - Iterative Problem Solving: The AI now iterates through possible solutions, refining its approach and learning from mistakes, leading to more accurate outputs. - Real-World Applications: It excels in fields like scientific research, coding, and even legal tech, providing tailored solutions to industries that require intricate reasoning capabilities - Enhanced Computational Processing: Though slower, the deliberate reasoning embedded in Strawberry results in significantly more complex problem-solving, with an emphasis on accuracy over speed. Strawberry introduces a new phase in AI-human collaboration. While previous models like GPT-4 offered impressive conversational abilities, Strawberry brings AI closer to the realm of thinking, moving away from simple word prediction to truly understanding the structure of complex problems. OpenAI’s goal is to make AI models more autonomous, capable of conducting deep research, solving scientific problems, and potentially making discoveries previously impossible with older models. More here: https://lnkd.in/eJdztrzd

3

OpenAI's #ChatGPT memory 🧠 is a fantastic feature, but it seems a bit... random. It remembers things here and there, and then suddenly, memory is full and it can't retain anything new. Here's what the memory function is: it allows ChatGPT to remember details between conversations to personalize responses better, adapting to your preferences and past discussions. But there are definitely a few areas for improvement. So, here's what I'd suggest to the ChatGPT PMs: - Give an explicit command to tell ChatGPT to remember specific things. - In the memory management window, allow users to select multiple items to forget. It would be so much faster than deleting one by one. Maybe these functionalities already exist? If so, someone please enlighten me💡. Always happy to learn! :-) #AI #ChatGPT #ProductFeedback #TechCommunity ________________________ ⚡ Love ❤️ this post? Add your comments 💭, re-share ♻️ with your network, follow me 🔔 for more insights, and DM me 📩 if you want to dive deeper into the complexity of drug discovery and R&D!

5

1 Kommentar

Thanks Thomas Hagemeijer for quoting me with the perspective on data analysis as the ultimate bottleneck of large-scale disease detection. The intersection of clinical data and life science technologies is where we will experience the biggest changes in medicine in the next few years. It's a bit of what we see at Web 4.0 at the frontier of human and machine - but focussing on the integration of clinical evidence and new technologies for the benefit of medicine and health. BigOmics Analytics Springboard Health Angels Singleron Biotechnologies H3 Health Hub BioRN Life Science Cluster Rhine-Neckar BioM Biotech Cluster Development GmbH Netzwerker BIODeutschland Sciomics

6

Jumping Into the Rabbit Holes of the Mind: Growth, Curiosity, and the Rule of 100 "Be brave enough to be bad at something new." A team member shared this quote recently, and it has been echoing in my mind since. It's a powerful reminder that starting something new can feel daunting, but here’s the secret: you’re never truly “bad.” When you step into uncharted territory, you bring your unique perspective, diverse experiences, and fresh approach. That uniqueness is your edge. Even if you’re venturing into unfamiliar terrain, your value lies in how differently you see the world compared to everyone else. Steven Bartlett captures this beautifully in his video, where he discusses how the most successful people aren’t necessarily the “best” at anything. Instead, they’re often masters at embracing learning, growth, and adapting. Being “good enough” in multiple areas can make you stand out far more than excelling in just one. https://lnkd.in/dqWfcGPz This brings us to the Rule of 100 (Jesse Itzler), a concept that truly resonates. It’s simple: invest 100 hours of deliberate practice into anything during a year—whether it’s improving a skill, building a habit, or chasing a goal—and you’ll be better than 95% of the world at the end of that year. You don’t need special talent, money, or connections, you need consistency. How empowering is it to understand that dedicating even just 16 minutes a day can deliver remarkable progress over time. The key is showing up, even when it’s hard, and trusting that small, consistent efforts lead to exponential growth. So here’s how I'm thinking about 2025 and my challenge for you: 💡 Jump into the rabbit holes of your mind. 💡 Explore that thing you’ve always wanted to try but hesitated because you feared being “bad.” 💡 Commit to just 16 minutes a day and see how far you can go. You don’t have to be great to start, but you do have to start to be great. #SelfImprovement #GrowthMindset #RuleOf100 #ContinuousLearning #BeBrave

30

3 Kommentare

Techbio Unicorn Owkin Unveils AI-Driven Oncology and Immunology Pipeline : 4 early-stage de novo drug discovery programs and 1 patent-protected in-licensed drug asset ! 🤖𝐀𝐈-𝐏𝐨𝐰𝐞𝐫𝐞𝐝 𝐏𝐢𝐩𝐞𝐥𝐢𝐧𝐞 Owkin's pipeline leverages several internal AI engines and multimodal patient data from 61 research centers. With a team of 110 data scientists, Owkin uses spatial multiomics data via MOSAIC, the world's largest dataset in oncology, to understand the tumor microenvironment. This expertise positions Owkin to tackle challenges in precision medicine. 🧬𝐀𝐈 𝐄𝐧𝐠𝐢𝐧𝐞𝐬 𝐒𝐞𝐥𝐞𝐜𝐭 𝐎𝐊𝐍𝟒𝟑𝟗𝟓 𝐟𝐨𝐫 𝐢𝐧𝐥𝐢𝐜𝐞𝐧𝐬𝐢𝐧𝐠 Owkin's biomarker engine created detailed EP2/EP4 signatures using multimodal patient data, identifying actionable biomarkers. The AI drug positioning engine screened over 30 cancer indications for OKN4395, identifying the most promising therapeutic combinations for clinical trial 👩‍⚕️𝐂𝐥𝐢𝐧𝐢𝐜𝐚𝐥 𝐓𝐫𝐢𝐚𝐥 𝐎𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐎𝐊𝐍𝟒𝟑𝟗𝟓 Owkin plans to use its clinical trial optimization solutions to bring OKN4395 to the patients most likely to benefit from this innovative therapeutic. By utilizing its unique AI-driven pipeline, Owkin aims to enhance precision medicine in oncology and immunology.success.

72

5 Kommentare

The world of #deeplearning and AI is assuredly augmenting human capability in the realm of science. Dev-ResNet (#devresnet) is a seminal deep learning AI model developed by researchers at the University of Plymouth. It is capable of accurately detecting and identifying key developmental milestones during embryonic stages by analyzing video recordings, rather than relying on static images. Why is this important you may ask. These deep learning models are pivotal and can pave the way for areas such as evolutionary studies and assessing impact of external factors, such as #climatechange, on human and animal development in the earliest stages of life! https://lnkd.in/geP4dQXk

4

A few thoughts on the impact of AI on Healthcare and a possible roadmap for the next 5 years #ai #healthcare #artificialintelligence https://lnkd.in/dUA-k4_b

1

𝐓𝐨 𝐁(𝐃) 𝐨𝐫 𝐧𝐨𝐭 𝐁(𝐃)? (𝐄𝐩𝐢𝐬𝐨𝐝𝐞 𝐈𝐕) You can’t drive the growth of a scientific company without the hard work of their scientists. They are the ones doing the hard work, generating the data and making the science believable. Any business strategy should be supported by well-defined (and sometimes crazy) scientific ideas. ________________________________________________________________________________ I find myself at crossroads in my career, exploring Business Development as my next step. As I dive into the world of commercial partnerships, negotiation skills and market trends, I’ll be sharing my reflections. Perhaps it will help someone else who is navigating a similar path. Interested in joining the conversation? Get in touch! _________________________________________________________________________________ #InnovationStrategy #BusinessGrowth #ScienceInBusiness #SuccessThroughScience

6

"Politics is always an interesting piece. From my reflection, a technology podcast serves two main purposes. First, it discusses technology for a general audience. Second, it talks about tech within the context of a techie podcast. As a tech podcast, you represent the techie customer persona who is not just interested in tech but also has similar thoughts or questions about global geopolitics, parenting, and other topics that, while not directly about tech, have a relevant angle. Politicians have a strong impact on the rules and regulations surrounding tech. So, I wouldn’t say no to having a politician with a strong point of view on technology come on the podcast." - Jeremy Au 区汉辉, Host of BRAVE Southeast Asia Tech Podcast Watch, listen or read the full insight at https://lnkd.in/g_Dhgdms Nonton, dengar atau baca wawasan lengkapnya di https://lnkd.in/gQmZZh2n 观看、收听或阅读全文，请访问 https://lnkd.in/gEum_MTC Xem, nghe hoặc đọc toàn bộ thông tin chi tiết tại https://lnkd.in/gc3avCz3 #SupplyChainDisruption #SingaporePort #Geopolitics #USChinaTradeWar #BidenTariffs #TrumpPolicies #ClimateChangeImpact #BeltAndRoad #BRAVEpodcast

1

If there's anything I've learned over the last 18 months I've been working at Synthace it's that if Markus Gershater is discussing drug discovery, I should listen. I encourage anyone and everyone to do the same and check out the latest "Data Analytics and AI Podcast" with Amar Drawid, Ph.D." below!

8

1 Kommentar

GenAI and Automation of Boring Tasks: Embracing AI Today AI is advancing at lightning speed 🚀, and while there are cutting-edge applications like #AlphaFold making headlines (and getting The Nobel Prize), most of AI value, today, lies in automating tasks that might seem mundane. Take Certara’s #CoAuthor or Insilico Medicine’s #Dora, for example. These AI tools are transforming the way scientists and researchers handle documentation and writing ✍️, freeing up time for higher-value tasks. It’s not that these tools aren’t groundbreaking—quite the opposite. If you’re not leveraging them, it’s like saying, “I have a candle 🕯️; I don’t need electricity ⚡️ and a light bulb 💡.” I recently heard a researcher say, “These AI writing tools write better than most of my colleagues.” This sentiment captures exactly where AI productivity is today. Sure, tools like CoAuthor may not feel as “fancy” as predictive models for ADME properties or toxicity, but they are the reality of AI-driven productivity right now. And we’ll get to the point where AI can reliably predict safety, toxicity, and clinical trial outcomes—but we’re not quite there yet. The key is to embrace the AI tools we have today, because those who belittle current AI productivity tools will likely resist future innovations too. Let’s not wait for the next big breakthrough to act. The tools we have now are already transforming the industry, and it’s only the beginning. The future is coming faster than you think, but the question is: are you ready to embrace what’s already here? _____________ Love ❤️ this post? Add your comments 💭, re-share ♻ it with your network, follow me 🔔 for more posts like this, and DM me 📩 if you want to deep dive on this topic!

7

For drug discovery scientists, consultants and vendors … Announcing targeted high-impact initiatives at HitchhikersAI + a review paper describing the challenges in the adoption of AI & Data.  Initiatives: ✅ Find bite-size solutions for your scientific problems ✅ Build relationship opportunities with biotechs, by consultants partnering with vendors ✅ A market penetration app (LLM-driven), to help vendors diagnose why they are struggling and to make actionable suggestions ✅ Going from data to quality decisions, fast ✅ Helping scientists cost-effectively construct decisions from large sets of research papers For more info on the initiatives: https://lnkd.in/epDWQuB4 Call to Action: ❓ [SCIENTISTS] Do you have a scientific problem where you need a bite-size solution? We may be able to find something for you. All we need is for you to send us your scientific problem. ❓ [VENDORS & CONSULTANTS] Would any of the initiatives above help you engage with your customers faster? These initiatives come with the support of our Steering Committee and our friends at 20/15 Visioneers and Zifo RnD Solutions.  Together, we will leverage our networks to ensure that participating vendors & consultants are able to engage with and get feedback from scientists. Join Us [Nothing to sign, Nothing to pay]: HitchhikersAI is a grass-roots no-cost community, seeking to accelerate drug discovery and development with AI & Data.  Please join our initiatives and shape them to directly help your scientific and business needs. Contact me to get involved: raminderpal@hitchhikersai.org SLAS (Society for Laboratory Automation and Screening)  ELRIG UK  Drug Target Review  Drug Discovery World  Sanj Kumar  Lesley Granberg  Vicki Loise  Pistoia Alliance  Mary Geismann  Boston Cambridge Biotech Networks  San Diego Biotechnology Network  San Francisco Biotech Networks  myBIO Community - Biotechnology connections myBIO Community - Biotechnology connections cvijic boris  Jesse Johnson  Anne Carpenter  Abhishek Jha  Catarina (Katie) Campbell  John Overington  Becky Upton BioIndustry Association (BIA) John F. Conway Neil Wilkie Andreas Bender Janette Thomas Andre Franca, PhD Nina Truter Sujeegar Jeevanandam Eli Pollock Jamie Cartland Jorge Cortell Tim Jarrett Charles Sinclair Graham Combe Sitta Sittampalam Tony Sedgwick https://lnkd.in/eicivKZf

40

Since the successful launch of ChatGPT, AI has taken center stage in technology discussions (goodbye, blockchain—it was nice knowing you for a moment). In the pharmaceutical sector, this topic is experiencing exponential growth, with the ongoing development of AI tools at all levels of the industry aimed at accelerating, optimizing, and expanding pharma business R&D and operations. This article highlights one area within the pharmaceutical sector that is reaping the benefits of AI development - drug discovery -, yet still needs to prove its worth. In my view, the key takeaway from this article—and the wider AI discourse—is encapsulated in this statement from Edward Tiam. 🌐 “Big promises have been made about how transformative this technology will be, and expectations have been high. But, though there have certainly been significant advancements, those expectations haven’t quite been met, and that is influencing the market” We all have high expectations for this technology, however, the challenge lies in managing those expectations and keeping realistic. While AI is transforming how we work, organize, produce, and analyze data, it still requires time to fully develop and in the highly regulated pharma field, too high expectations may lead to an unnecessary business bubble.  #aipharma #artificialintelligence #aihealthcare https://lnkd.in/ddKST489

8

1 Kommentar

Interesting discussion at “AI/ML is Transforming R&D and Business” panel at BIO International Convention - led by Maude Tessier with Karen Akinsanya Joseph Mcdonald Jiye Shi and Frank Grams. What is the value of AI/ML = one way to assess: ask whether decisions made by a drug discovery/development project team with the help of genAI or physics based approaches/models enhance the probability of success, accelerate, improve outcomes. The data we use for training models is only limited to the small fraction of targets/problems what we have tried to address. The more important data sets are the additional ones being generated to allow us to target the targets that haven’t been addressed previously.

29

Generative AI's power continues to deliver surprisingly compelling results. We are way past the Turing Test. Check out this article from #jama. https://lnkd.in/d_7X23yK LLMs generated convincing medical research abstracts indistinguishable from surgeons as evident in this cross-sectional study comparing 10 medical abstracts written by surgical trainees and senior surgeons or generated by LLMs. Blinded expert surgeon-reviewers were asked to grade and rank these abstracts. There was no statistical difference in the grades or ranks of abstracts generated by the language model when compared with abstracts written by surgical trainees or senior surgeons. These findings suggest that when appropriately trained with background literature, abstract formatting, primary research data, and a thorough prompt, chatbots can generate medical research abstracts that are difficult to distinguish from surgeon-scientist–written abstracts. The AI journey continues.

34

I feel seen! The authors articulated the problem well, though I feel as though the solutions are quite weak. “Cross-functional collaboration champion?” Feels like corporate BS… if a company really wants to support and encourage cross-functional collaboration, set the budgets and decision-making cross-functionally. Dont just recognize the emotional labor done by individuals that transition into and out of different silo cultures on an hourly basis ..disproportionately incentivize and reward that emotional labor. When looking for leaders to take the next big project / initiative / role, don’t focus on years of experience in a particular function, focus on ability to learn, grow, and adapt quickly and ability to align the right stakeholders to achieve a goal. Im sure theres so much more that companies can do - what else can you think of?

34

3 Kommentare