News

The New York Times is suing Microsoft and OpenAI. What is this about? Open AI and Microsoft have been in negotiations for several months over proper payment for the use of articles published by the New York Times and at this point, it looks as if those negotiations have broken down. As I am sure you know Chatgpt, the main product from OpenAI, scrapes the web for data and stories to use in their training data. Other organizations such as Gannett, News Corp, and the Wall Street Journal are still in negotiations for the use of their articles in training the AI tools created by Open AI.

Prompt

One way I have found to prompt Midjourney to create a more original painting is by suggesting a style. This can be in the style of Cezanne, Miro, Le Corbusier, and so on. Mondrian has become one of my favorite artist styles. Another way is by suggesting a painting medium as in watercolor.

This is an image I used in a recent blog post on Pluto.

Path to Explainability, ptr 3 Healthcare

In our first edition, we explored the critical link between explainability and trust in AI, emphasizing the importance of clear explanations in helping users understand AI systems' capabilities and limitations. The second edition delved into key considerations for explaining AI systems, discussing the importance of calibrating user trust, trust calibration throughout the user experience, optimizing for understanding, and managing AI's influence on user decisions. We also examined the factors contributing to user trust in AI, focusing on ability, reliability, and benevolence.

Health Care AI for Diagnosis

Explainable AI (XAI) has significant potential in the healthcare industry, providing vital transparency and understanding to AI systems across various applications.

One area is performing sanity checks on AI models under development to ensure accurate functioning before validation for clinical use. For example, researchers at Mount Sinai Hospital created a model to identify high-risk patients from X-ray images. However, they later found the model was basing decisions on metadata from a specific X-ray machine rather than actual clinical data. XAI methods would have uncovered this issue earlier.

XAI can also help detect biases that may unfairly skew AI predictions. Identifying and mitigating discrimination is crucial for equitable healthcare. If an AI system was found to provide less accurate diagnoses for minority groups, XAI could pinpoint where corrective measures are needed.

In cases of disagreement between an AI and human experts, XAI delivers insights into how different variables influence the AI’s final recommendations. This allows clinicians to better critique and challenge the system’s conclusions. Building understanding facilitates trust in AI tools among medical professionals, leading to greater adoption of assistive AI decision-making.

Likewise, explaining personalized medicine recommendations enables more tailored and effective treatment plans for patients. XAI reveals why genetic, clinical, and other factors lead an AI to suggest specific therapies for individuals.

Across applications like clinical decision support and predictive healthcare modeling, transparent AI augments human expertise rather than replaces it. Elucidating how conclusions are reached empowers medical professionals to make more informed choices when utilizing AI assistance. Overall, implementing explainable AI wisely promotes safe, ethical progress in bringing advanced technology into the health sector.