AI Research
Microsoft and OpenAI reach agreement on… something – Computerworld
“OpenAI’s decision to recast its for-profit arm as a public benefit corporation while keeping control in the hands of its nonprofit parent is without precedent at this scale,” said Sanchit Vir Gogia, chief analyst and CEO at Greyhound Research. “It is a structure that fuses two competing logics: the relentless need to raise capital for models that cost billions to train, and the equally strong pressure from regulators, investors, and the public to demonstrate accountability.”
The proposed structure enables OpenAI to attract traditional investors who expect potential returns, while the nonprofit parent can ensure safety considerations aren’t sacrificed for profit. However, Gogia warned the hybrid model is “as much a gamble as it is an innovation,” noting concerns about “who carries liability if something goes wrong, whether fiduciary duty to investors will override social commitments when revenues are threatened.”
Charlie Dai, VP and principal analyst at Forrester, said the structure “could influence others because it balances capital access with mission-driven oversight” but warned that “regulatory scrutiny, lawsuits, and governance complexity introduce uncertainty around decision-making speed and long-term stability.”
Bayer is a global life sciences company operating across Pharmaceuticals, Consumer Health, and Crop Science. In fiscal 2024, the group reported €46.6 billion in sales and 94,081 employees, a scale that makes internal AI deployments consequential for workflow change and ROI.
The company invests heavily in research, with more than €6 billion allocated to R&D in 2024, and its leadership frames AI as an enabler for both sustainable agriculture and patient-centric medicine. Bayer’s own materials highlight AI’s role in planning and analyzing clinical trials as well as accelerating crop protection discovery pipelines.
This article examines two mature, internally used applications that convey the central role AI plays in Bayer’s core business goals:
- Herbicide discovery in crop science: Applying AI to narrow down molecular candidates and identify new modes of action.
- Clinical trial analytics in pharmaceuticals: Ingesting heterogeneous trial and device data to accelerate compliant analysis.
AI-Assisted Herbicide Discovery
Weed resistance is a mounting global challenge. Farmers in the US and Brazil are facing species resistant to multiple herbicide classes, driving up costs and threatening crop yields. Traditional herbicide discovery is slow — often 12 to 15 years from concept to market — and expensive, with high attrition during early screening.
Bayer’s Crop Science division has turned to AI to help shorten these timelines. Independent reporting notes Bayer’s pipeline includes Icafolin, its first new herbicide mode of action in decades, expected to launch in Brazil in 2028, with AI used upstream to accelerate the discovery of new modes of action.
Reuters reports that Bayer’s approach uses AI to match weed protein structures with candidate molecules, compressing the early discovery funnel by triaging millions of possibilities against pre-determined criteria. Bayer’s CropKey overview describes a profile-driven approach, where candidate molecules are designed to meet safety, efficacy, and environmental requirements from the start.
The company claims that CropKey has already identified more than 30 potential molecular targets and validated over 10 as entirely new modes of action. These figures, while promising, remain claims until independent verification.
For Bayer’s discovery scientists, AI-guided triage changes workflows by:
- Reducing early-stage wet-lab cycles by focusing on higher-probability matches between proteins and molecules.
- Integrating safety and environmental criteria into the digital screen, filtering out compounds unlikely to meet regulatory thresholds.
- Advancing promising molecules sooner, enabling earlier testing and potentially compressing development timelines from 15 years to 10.
Coverage by both Reuters and the Wall Street Journal notes this strategy is expected to reduce attrition and accelerate discovery-to-commercialization timelines.
The CropKey program has been covered by multiple independent outlets, a signal of maturity beyond a single press release. Reuters reports Bayer’s assertion that AI has tripled the number of new modes of action identified in early research compared to a decade ago.
The upcoming Icafolin herbicide, expected for commercial release in 2028, demonstrates that CropKey outputs are making their way into the regulatory pipeline. The presence of both media scrutiny and near-term launch candidates suggests CropKey is among Bayer’s most advanced AI deployments.
Video explaining Bayer’s CropKey process in crop protection discovery. (Source: Bayer)
By focusing AI on high-ROI bottlenecks in research and development, Bayer demonstrates how machine learning can trim low-value screening cycles, advancing only the most promising candidates into experimental trials. At the same time, acceleration figures reported by the company should be treated as claims until they are corroborated across multiple seasons, geographies, and independent trials.
Clinical Trial Analytics Platform (ALYCE)
Pharmaceutical development increasingly relies on complex data streams: electronic health records (EHR), site-based case report forms, patient-reported outcomes, and telemetry from wearables in decentralized trials. Managing this data volume and variety strains traditional data warehouses and slows regulatory reporting.
Bayer developed ALYCE (Advanced Analytics Platform for the Clinical Data Environment) to handle this complexity. In a PHUSE conference presentation, Bayer engineers describe the platform as a way to ingest diverse data, ensure governance, and deliver analytics more quickly while maintaining compliance.
The presentation describes ALYCE’s architecture as using a layered “Bronze/Silver/Gold” data lake approach. An example trial payload included approximately 300,000 files (1.6 TB) for 80 patients, requiring timezone harmonization, device ID mapping, and error handling before data could be standardized to SDTM (Study Data Tabulation Model) formats. Automated pipelines provide lineage, quarantine checks, and notifications. These technical details were presented publicly to peers, reinforcing their credibility beyond internal marketing.
For statisticians and clinical programmers, ALYCE claims to:
- Standardize ingestion across structured (CRFs), semi-structured (EHR extracts), and unstructured (device telemetry) sources.
- Automate quality checks through pipelines that reduce manual intervention and free staff up to focus on analysis.
- Enable earlier insights by preparing analysis-ready datasets faster, shortening the lag between data collection and review.
These objectives are consistent with Bayer’s broader statement that AI is being used to plan and analyze clinical trials safely and efficiently.
PHUSE is a respected industry forum where sponsors share methods with peers, and Bayer’s willingness to disclose technical details indicates ALYCE is in production. While Bayer has not released precise cycle-time savings, its emphasis on elastic storage, regulatory readiness, and speed suggests measurable efficiency gains.
Given the specificity of the presentation — real-world payloads, architecture diagrams, and validation processes — ALYCE appears to be a mature platform actively supporting Bayer’s clinical trial programs.
Screenshot from Bayer’s PHUSE presentation illustrating ALYCE’s automated ELTL pipeline.
(Source: PHUSE)
Bayer’s commitment to ALYCE reflects its broader effort to modernize and scale clinical development. By consolidating varied data streams into a single, automated environment, the company positions itself to shorten study timelines, reduce operational overhead, and accelerate the movement of promising therapies from discovery to patients. This infrastructure also prepares Bayer to expand AI-driven analytics across additional therapeutic areas, supporting long-term competitiveness in a highly regulated industry.
While Bayer has not published specific cycle-time reductions or quantified cost savings tied directly to ALYCE, the company’s willingness to present detailed payload volumes and pipeline architecture at PHUSE indicates that the platform is actively deployed and has undergone peer-level scrutiny. Based on those disclosures and parallels with other pharma AI implementations, reasonable expectations include faster data review cycles, earlier anomaly detection, and improved compliance readiness. These outcomes—though not yet publicly validated—suggest ALYCE is reshaping Bayer’s trial workflows in ways that could yield significant long-term returns.
AI Research
The STARD-AI reporting guideline for diagnostic accuracy studies using artificial intelligence
Reichlin, T. et al. Early diagnosis of myocardial infarction with sensitive cardiac troponin assays. N. Engl. J. Med. 361, 858–867 (2009).
Hawkes, N. Cancer survival data emphasise importance of early diagnosis. BMJ 364, l408 (2019).
Neal, R. D. et al. Is increased time to diagnosis and treatment in symptomatic cancer associated with poorer outcomes? Systematic review. Br. J. Cancer 112, S92–S107 (2015).
Leifer, B. P. Early diagnosis of Alzheimer’s disease: clinical and economic benefits. J. Am. Geriatr. Soc. 51, S281–S288 (2003).
Crosby, D. et al. Early detection of cancer. Science 375, eaay9040 (2022).
Fleming, K. A. et al. The Lancet Commission on diagnostics: transforming access to diagnostics. Lancet 398, 1997–2050 (2021).
Whiting, P. F., Rutjes, A. W., Westwood, M. E. & Mallett, S. A systematic review classifies sources of bias and variation in diagnostic test accuracy studies. J. Clin. Epidemiol. 66, 1093–1104 (2013).
Glasziou, P. et al. Reducing waste from incomplete or unusable reports of biomedical research. Lancet 383, 267–276 (2014).
Ioannidis, J. P. et al. Increasing value and reducing waste in research design, conduct, and analysis. Lancet 383, 166–175 (2014).
Lijmer, J. G. et al. Empirical evidence of design-related bias in studies of diagnostic tests. JAMA 282, 1061–1066 (1999).
Irwig, L., Bossuyt, P., Glasziou, P., Gatsonis, C. & Lijmer, J. Designing studies to ensure that estimates of test accuracy are transferable. BMJ 324, 669–671 (2002).
Moons, K. G., van Es, G. A., Deckers, J. W., Habbema, J. D. & Grobbee, D. E. Limitations of sensitivity, specificity, likelihood ratio, and Bayes’ theorem in assessing diagnostic probabilities: a clinical example. Epidemiology 8, 12–17 (1997).
Bossuyt, P. M. et al. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Ann. Intern. Med. 138, W1–W12 (2003).
Bossuyt, P. M. et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 351, h5527 (2015).
Cohen, J. F. et al. STARD 2015 guidelines for reporting diagnostic accuracy studies: explanation and elaboration. BMJ Open 6, e012799 (2016).
Cohen, J. F. et al. STARD for Abstracts: essential items for reporting diagnostic accuracy studies in journal or conference abstracts. BMJ 358, j3751 (2017).
Korevaar, D. A. et al. Reporting diagnostic accuracy studies: some improvements after 10 years of STARD. Radiology 274, 781–789 (2015).
Korevaar, D. A., van Enst, W. A., Spijker, R., Bossuyt, P. M. & Hooft, L. Reporting quality of diagnostic accuracy studies: a systematic review and meta-analysis of investigations on adherence to STARD. Evid. Based Med. 19, 47–54 (2014).
Miao, Z., Humphreys, B. D., McMahon, A. P. & Kim, J. Multi-omics integration in the age of million single-cell data. Nat. Rev. Nephrol. 17, 710–724 (2021).
Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209 (2018).
Williamson, E. J. et al. Factors associated with COVID-19-related death using OpenSAFELY. Nature 584, 430–436 (2020).
Lu, R. et al. Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding. Lancet 395, 565–574 (2020).
De Fauw, J. et al. Clinically applicable deep learning for diagnosis and referral in retinal disease. Nat. Med. 24, 1342–1350 (2018).
McKinney, S. M. et al. International evaluation of an AI system for breast cancer screening. Nature 577, 89–94 (2020).
Topol, E. J. High-performance medicine: the convergence of human and artificial intelligence. Nat. Med. 25, 44–56 (2019).
Benjamens, S., Dhunnoo, P. & Meskó, B. The state of artificial intelligence-based FDA-approved medical devices and algorithms: an online database. NPJ Digit. Med. 3, 118 (2020).
Liu, X. et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit. Health 1, e271–e297 (2019).
Liu, X. et al. Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nat. Med. 26, 1364–1374 (2020).
Rivera, S. C., Liu, X., Chan, A.-W., Denniston, A. K. & Calvert, M. J. Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI Extension. BMJ 370, m3210 (2020).
Collins, G. S. et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ 385, e078378 (2024).
Tejani, A. S. et al. Checklist for Artificial Intelligence in Medical Imaging (CLAIM): 2024 Update. Radiol. Artif. Intell. 6, e240300 (2024).
Aggarwal, R. et al. Diagnostic accuracy of deep learning in medical imaging: a systematic review and meta-analysis. NPJ Digit.Med. 4, 65 (2021).
McGenity, C. et al. Artificial intelligence in digital pathology: a systematic review and meta-analysis of diagnostic test accuracy. NPJ Digit. Med. 7, 114 (2024).
Rajkomar, A. et al. Scalable and accurate deep learning with electronic health records. NPJ Digit. Med. 1, 18 (2018).
Moons, K. G. M., de Groot, J. A. H., Linnet, K., Reitsma, J. B. & Bossuyt, P. M. M. Quantifying the added value of a diagnostic test or marker. Clin. Chem. 58, 1408–1417 (2012).
Bossuyt, P. M. M., Reitsma, J. B., Linnet, K. & Moons, K. G. M. Beyond diagnostic accuracy: the clinical utility of diagnostic tests. Clin. Chem. 58, 1636–1643 (2012).
Gallifant, J. et al. The TRIPOD-LLM reporting guideline for studies using large language models. Nat. Med. 31, 60–69 (2025).
Kelly, C. J., Karthikesalingam, A., Suleyman, M., Corrado, G. & King, D. Key challenges for delivering clinical impact with artificial intelligence. BMC Med. 17, 195 (2019).
Yang, Y., Zhang, H., Gichoya, J. W., Katabi, D. & Ghassemi, M. The limits of fair medical imaging AI in real-world generalization. Nat. Med. 30, 2838–2848 (2024).
The White House. Delivering on the Promise of AI to Improve Health Outcomes. https://bidenwhitehouse.archives.gov/briefing-room/blog/2023/12/14/delivering-on-the-promise-of-ai-to-improve-health-outcomes/ (2023).
Coalition for Health AI. Blueprint for Trustworthy AI Implementation Guidance and Assurance for Healthcare. https://www.chai.org/workgroup/responsible-ai/blueprint-for-trustworthy-ai (2023).
Guni, A., Varma, P., Zhang, J., Fehervari, M. & Ashrafian, H. Artificial intelligence in surgery: the future is now. Eur. Surg. Res. https://doi.org/10.1159/000536393 (2024).
Chen, R. J. et al. Algorithmic fairness in artificial intelligence for medicine and healthcare. Nat. Biomed. Eng. 7, 719–742 (2023).
Krakowski, I. et al. Human-AI interaction in skin cancer diagnosis: a systematic review and meta-analysis. NPJ Digit. Med. 7, 78 (2024).
Moor, M. et al. Foundation models for generalist medical artificial intelligence. Nature 616, 259–265 (2023).
Tu, T. et al. Towards generalist biomedical AI. NEJM AI 1, AIoa2300138 (2024).
Acosta, J. N., Falcone, G. J., Rajpurkar, P. & Topol, E. J. Multimodal biomedical AI. Nat. Med. 28, 1773–1784 (2022).
Barata, C. et al. A reinforcement learning model for AI-based decision support in skin cancer. Nat. Med. 29, 1941–1946 (2023).
Mankowitz, D. J. et al. Faster sorting algorithms discovered using deep reinforcement learning. Nature 618, 257–263 (2023).
Corso, G., Stark, H., Jegelka, S., Jaakkola, T. & Barzilay, R. Graph neural networks. Nat. Rev. Methods Primers 4, 17 (2024).
Li, H. et al. CGMega: explainable graph neural network framework with attention mechanisms for cancer gene module dissection. Nat. Commun. 15, 5997 (2024).
Pahud de Mortanges, A. et al. Orchestrating explainable artificial intelligence for multimodal and longitudinal data in medical imaging. NPJ Digit. Med. 7, 195 (2024).
Johri, S. et al. An evaluation framework for clinical use of large language models in patient interaction tasks. Nat. Med. 31, 77–86 (2025).
EQUATOR Network. Enhancing the QUAlity and Transparency Of health Research. https://www.equator-network.org/
Sounderajah, V. et al. Developing specific reporting guidelines for diagnostic accuracy studies assessing AI interventions: the STARD-AI Steering Group. Nat. Med. 26, 807–808 (2020).
Sounderajah, V. et al. Developing a reporting guideline for artificial intelligence-centred diagnostic test accuracy studies: the STARD-AI protocol. BMJ Open 11, e047709 (2021).
AI Research
This Artificial Intelligence (AI) Stock Will Beat Opendoor Technologies over the Next 3 Years
-
Opendoor stock has jumped more than 1,000% in the last three months.
-
Upstart has a number of similarities to Opendoor.
-
The fintech company has proven its model can work even in a high-interest-rate environment.
Opendoor Technologies (NASDAQ: OPEN) dazzled investors over the last three months like few other stocks. The online home-flipper jumped an incredible 1,400% over the last three months, going from a little over $0.50 a share to more than $10 at one point.
The rally began with hedge-fund manager Eric Jackson making the case that the stock could be the next Carvana, which jumped to almost 100 times its original price after nearly going bankrupt in 2022. That argument gained steam online and helped turn Opendoor into a meme stock, as it initially surged on high volume and no news.
Since then, the stock gained on real news. That includes the prospect of the Federal Reserve lowering interest rates next week and later in the year, and the company’s board overhauling its management team. In August, embattled CEO Carrie Wheeler stepped down; after hours on Wednesday, Opendoor named Shopify chief operating officer Kaz Nejatian as its new CEO, which sent the stock up 80% on Thursday.
Additionally, the company said that co-founders Keith Rabois and Eric Wu were rejoining the board of directors, and ventures associated with them were investing $40 million into Opendoor. It’s easy to see how that news would inject enthusiasm into the stock, especially after it was on the verge of being delisted by the Nasdaq stock exchange earlier.
However, nothing’s really changed for Opendoor as a business in the last three months. The company never reported a full-year profit, and the business is expected to shrink this quarter due to the weak housing market.
It’s still a high risk with a questionable business model. If you’re looking for a similar stock that can capitalize on falling interest rates, I think that Upstart Holdings (NASDAQ: UPST) is a better bet, and that it can outperform Opendoor over the next three years.
Upstart has a number of things in common with Opendoor. Both went public around the same time in 2020, and initially surged out of the gate before plunging in 2022 as interest rates rose and tech stocks crashed.
Upstart is a loan originator. It uses artificial intelligence (AI) technology to screen applicants, producing results it claims are significantly better than traditional FICO scores. Once it creates a loan, it typically sells it to one of its funding partners, so it doesn’t keep the debt on its books.
-
Business2 weeks agoThe Guardian view on Trump and the Fed: independence is no substitute for accountability | Editorial
-
Tools & Platforms1 month ago
Building Trust in Military AI Starts with Opening the Black Box – War on the Rocks
-
Ethics & Policy2 months agoSDAIA Supports Saudi Arabia’s Leadership in Shaping Global AI Ethics, Policy, and Research – وكالة الأنباء السعودية
-
Events & Conferences4 months agoJourney to 1000 models: Scaling Instagram’s recommendation system
-
Jobs & Careers3 months agoMumbai-based Perplexity Alternative Has 60k+ Users Without Funding
-
Podcasts & Talks2 months agoHappy 4th of July! 🎆 Made with Veo 3 in Gemini
-
Education3 months agoVEX Robotics launches AI-powered classroom robotics system
-
Education2 months agoMacron says UK and France have duty to tackle illegal migration ‘with humanity, solidarity and firmness’ – UK politics live | Politics
-
Podcasts & Talks2 months agoOpenAI 🤝 @teamganassi
-
Funding & Business3 months agoKayak and Expedia race to build AI travel agents that turn social posts into itineraries