AI Safety on the Chopping Block to eliminate oversight that could slow down or restrict Musk's and others' AI ambitions

 by Mary Harrsch © 2025

"The US AI Safety Institute was created to address a wide range of concerns associated with the rapid advancement of artificial intelligence technologies. Its primary mission is to conduct rigorous research into the potential risks of AI, from issues of algorithmic bias and cybersecurity vulnerabilities to long-term existential threats posed by superintelligent systems. By funding interdisciplinary research, fostering collaboration between academia, industry, and government, and advising policymakers on regulation and ethical standards, the Institute is intended to serve as a national hub for ensuring that AI development proceeds in a safe, transparent, and responsible manner."

https://www.hpbl.co.in/market/us-ai-safety-institute-could-face-big-cuts-implications-challenges-and-future-prospects/

Although many of you tend to roll your eyes whenever I mention artificial intelligence and my opinion that it will enhance much of our research even in the humanities, I think this announcement is extremely troubling and bodes ill for all of us.

The perceived risks of advanced artificial intelligence (AI) span multiple domains, including ethical, societal, security, and existential concerns. Some of the most significant risks include:

1. Misinformation & Manipulation

AI-generated deepfakes and text-based misinformation can be used to manipulate public opinion, elections, and markets.

Large language models can spread biased or misleading information, even unintentionally.

2. Bias & Discrimination

AI systems trained on biased data can perpetuate and amplify discrimination in hiring, lending, healthcare, and law enforcement.

Lack of transparency in AI decision-making makes it difficult to address biases.

3. Job Displacement & Economic Disruption

Automation of white-collar and blue-collar jobs may lead to mass unemployment, particularly in industries like customer service, software development, and transportation. Economic inequality could widen as AI benefits a small number of corporations and individuals.

4. Loss of Human Oversight & Accountability

AI decision-making can become too complex for humans to understand, making it difficult to hold systems accountable for errors or harmful actions. Autonomous weapons and AI-driven military systems could make lethal decisions without human intervention.

5. Cybersecurity & AI-powered Attacks

AI can be exploited to conduct highly sophisticated cyberattacks, including automated hacking, phishing, and misinformation campaigns. AI-driven social engineering could make scams and fraud more convincing.

[For a glimpse of what AI-powered attacks look like, check out the new Netflix limited series "Zero Day" starring Robert De Niro"]

6. Superintelligence & Existential Risk

Some experts fear that AI could surpass human intelligence and act in ways that are unpredictable or uncontrollable. An AI system optimizing for the wrong goal (e.g., maximizing efficiency without ethical considerations) could have catastrophic consequences.

7. Privacy Violations & Mass Surveillance

AI-driven surveillance can erode privacy and civil liberties, allowing governments and corporations to track individuals in unprecedented detail. AI-powered facial recognition and predictive policing can lead to over-policing and discrimination.

8. AI Alignment Problem

Ensuring that AI systems follow human values and ethical guidelines is an ongoing challenge. A misaligned AI system could pursue harmful objectives if its goals are not carefully defined and constrained.

9. Dependence on AI & Loss of Critical Thinking

Overreliance on AI in decision-making could weaken human problem-solving skills and critical thinking. Societies may become overly dependent on AI-driven automation, leading to vulnerabilities if AI systems fail.

10. Weaponization of AI

Autonomous drones and AI-driven cyber warfare could increase the speed and scale of conflicts. AI could be used to create novel biological or chemical weapons. While AI offers significant benefits, these risks highlight the importance of robust oversight, ethical AI development, and international cooperation to mitigate potential harms.

Since the proposed cuts are being directed by Elon Musk, whose own AI development platform xAI is in heated competition with ChatGPT, ClaudeAI, and others, I felt compelled to explore Musk's collision with AI development oversight agencies to see if he is willing to destroy this organization to free his own company from any government compliance so he can speed his own company's development and generate more profit with total disregard for the ultimate expense in loss of freedom, gross exploitation, and the quality of our lives to millions of us worldwide.

As I suspected, several issues have been raised regarding Musk's AI operations:

Risk Management Practices: A study conducted by the French nonprofit SaferAI, published in October 2024, evaluated the risk management protocols of leading AI companies. xAI received the lowest possible score, indicating significant deficiencies in its risk management practices. The study highlighted concerns about the company's preparedness to handle potential AI-related risks.

AI Model Vulnerabilities: Recent research by AI security firm Adversa AI identified substantial cybersecurity vulnerabilities in xAI's Grok 3 model. These findings suggest that the model could be susceptible to exploitation, raising questions about the robustness of xAI's AI systems.

Environmental Compliance: In August 2024, environmental advocates accused xAI of operating natural gas turbines at its Memphis data center without the necessary permits. These turbines, with a combined capacity of approximately 100 MW, were reportedly emitting pollutants that could worsen local air quality. The Southern Environmental Law Center formally requested investigations by the Environmental Protection Agency and local health authorities.

ChatGPT points out:

"Cutting funding to the U.S. AI Safety Institute would directly benefit Elon Musk and his company xAI, as well as other AI developers who want to avoid regulatory scrutiny. The AI Safety Institute was established to evaluate and mitigate the risks posed by advanced AI models, including cybersecurity vulnerabilities, misinformation, and potential misuse. If this oversight is weakened, companies like xAI would face fewer regulatory hurdles and could deploy AI models with less accountability and fewer safety checks.

Given that:

xAI scored poorly in AI risk management (per the SaferAI study),

Grok 3 was found to have significant security vulnerabilities, and

Musk has a history of resisting regulations in other ventures (Tesla, SpaceX, Neuralink), his push to "save money" by defunding AI safety efforts looks less like fiscal responsibility and more like a move to eliminate oversight that could slow down or restrict his AI ambitions.

To me, one of the most dangerous threats from AI development without human oversight is described in Risk No. 6 above - "An AI system optimizing for the wrong goal (e.g., maximizing efficiency without ethical considerations) could have catastrophic consequences." Maybe Musk has already delegated his role in maximizing efficiency to his flawed xAI and we are already suffering the consequences!

Kyoto of the 1930s and the 1980s

 by Mary Harrsch © 2025

I happen to see this interesting video about the Japanese artist Toshi Yoshida entitled "Walking through Kyoto in the 1930s" showcasing his paintings:

I thoroughly enjoyed my visit to Kyoto in 1982. Even though western-style dress was widely accepted, I still saw older people wearing traditional clothing. At one temple I asked an elderly gentleman who was walking with his granddaughter If I could take his picture and he agreed. He was dressed in a dark navy blue (known as "kon-iro") montsuki kimono (formal winter kimono) with hakama (divided skirt-like pants) for formal occasions.) His granddaughter, though, was wearing western-style T-shirt and shorts.

Here is my picture of the Japanese grandfather and his granddaughter feeding the pigeons at a temple in Kyoto in October 1982. I was the only female member of the first alfalfa products trade team to Japan and Korea. We had a day off so I caught the bullet train from Tokyo to Kyoto and spent a wonderful day there. I took one of those tours that took me to a number of temples so I'm not sure which one this was although it could have been Kiyomizudera (Pure Water Temple) as I faintly remember a fountain with an ornate dragon.

Grandfather and granddaughter feeding pigeons at a temple in Kyota in October 1982. 

I uploaded my picture to ClaudeAI and it observed:

"Given the timing (October) and color (dark navy blue), this was most likely a montsuki kimono or a formal winter kimono. Since you mention it was an elderly gentleman, it makes even more sense as older Japanese men were more likely to wear traditional formal attire in the 1980s.

The dark navy color (known as "kon-iro" in Japanese) is a traditional, dignified choice that was and still is very common for men's formal kimonos. October in Japan is moving into cooler weather, so it would have been appropriate to wear a lined silk kimono rather than a lighter summer yukata.

What's particularly interesting about your timing - 1982 was a period when traditional kimono wearing was already becoming less common in everyday life in Japan, so capturing that image of an elderly gentleman in traditional dress would have documented an important cultural moment.

Men's traditional Japanese clothing can be called kimono, though there are some specific terms and distinctions worth noting:

The general term "kimono" (着物) applies to both men's and women's traditional garments. However, men's kimonos tend to be more subdued in color and pattern compared to women's, and they're typically worn with hakama (袴, divided skirt-like pants) for formal occasions.

Some specific types of men's kimono include:

Montsuki - A formal black kimono with family crests

Haori - A hip-length jacket worn over the kimono

Yukata - A casual, unlined summer kimono typically made of cotton

The way of wearing a kimono also differs between men and women. Men's kimonos have shorter sleeves, and the obi (belt) is tied differently and is much simpler than women's elaborate obi styles.

Contacting your Congressional legislators urging them to stop Executive Overreach

 As a retired faculty member of the University of Oregon and a voter for the past 57 years, I remain committed to our democratic system, where all citizens are treated with equality and respect under the rule of law. However, after weeks of chaos and blatant constitutional violations by the executive branch, I am deeply concerned by Congress’s lack of action in restoring its legislative authority, particularly over national security, financial oversight, and civil service protections.

So I worked with artificial intelligence tools to develop a list of specific actions Congress can take to restore its constitutional role and block harmful overreach by the executive branch. I sent this to my senators and Congresswoman but I could not send letters to anyone else's representatives because their online contact form will not let you proceed when it verifies your address. So I'm sharing my letter here and urging you all to feel free to use it or any part of it to contact your representatives. The letter mentions legal actions your state Attorney Generals can take too, so you can copy and paste those sections and send them to your state AGs as well.

Dear [Legislator’s Name],

I remain committed to our democratic system, where all citizens are treated with equality and respect under the rule of law. However, after weeks of chaos and blatant constitutional violations by the executive branch, I am deeply concerned by Congress’s lack of action in restoring its legislative authority, particularly over national security, financial oversight, and civil service protections.

Below, I outline specific actions Congress can take to restore its constitutional role and block harmful overreach by the executive branch.

---

1. Block Unauthorized Access to Government Records

Legislative Actions:

• Introduce/co-sponsor bills strengthening penalties for unauthorized access.
• Improve security clearance procedures via amendments to the Federal Information Security Management Act (FISMA).
• Include provisions in appropriations bills mandating enhanced security protocols.

Oversight Powers:

• Call emergency hearings in the House Financial Services or Senate Banking Committees.
• Request investigations by the Government Accountability Office (GAO).
• Demand briefings from agency heads regarding security clearance compliance.

Administrative Pressure:

• Write formal letters demanding explanations from agency heads.
• Request Inspector General (IG) investigations into specific breaches.
• Push for immediate implementation of stricter access controls.

---

2. Prevent Unsubstantiated Executive-Imposed Tariffs

Legislative Actions:

• Pass bills revoking/restricting presidential tariff authority under Section 232 and Section 301 of trade laws.
• Amend trade laws to require congressional approval for specific tariff actions.
• Restrict funds for tariff implementation through appropriations bills.

Oversight & Judicial Engagement:

• Hold hearings documenting economic/security impacts via Ways & Means and Senate Finance Committees.
• File amicus briefs supporting state/business litigation against harmful tariffs.
• Encourage state attorneys general (AGs) to challenge tariffs that disproportionately impact state economies.

Key Legal Precedents:

• Massachusetts v. EPA (2007) – State standing in federal cases.
• Wyoming v. Oklahoma (1992) – Economic harm as grounds for legal action.
• Administrative Procedure Act – Challenge inadequate cost-benefit analyses.

---

3. Restore Critical Public Communications from Federal Agencies

Legislative & Oversight Actions:

• Mandate public health and safety disclosures from the CDC, NTSB, FAA, NOAA, EPA, and FDA, ensuring that such information is made publicly available to all news outlets and not restricted to any single platform.
• Use appropriations bills to condition funding on transparency.
• Request GAO investigations into communication restrictions.

State-Level & Administrative Options:

• Ensure State health departments can maintain independent reporting.
• Ensure Federal employees can use whistleblower protections to report suppression.
• Invoke Scientific Integrity Policies to protect public information access. These existing policies ensure that government scientists can conduct and share research free from political interference, safeguarding transparency in public health, environmental data, and safety regulations. Key policies include the White House OSTP Memorandum (2021), 42 CFR Part 93 (Public Health Service Policies on Research Misconduct), EPA Order 3120.5 (EPA Scientific Integrity Policy), NAO 202-735D (NOAA Scientific Integrity Policy), and 305 DM 3 (Department of the Interior Scientific Integrity Policy). These policies ensure that government scientists can conduct and share research free from political interference, safeguarding transparency in public health, environmental data, and safety regulations.

---

4. Protect Civil Servants from Unlawful Termination

Congressional Oversight:

• Call emergency hearings, subpoena agency heads.
• Demand IG reviews of federal personnel actions.

Legislative Actions:

• Strengthen whistleblower protections through new legislation.
• Add job security provisions in appropriations bills.
• File amicus briefs supporting civil servant lawsuits.

---

5. Limit the President’s Military Authority

Strengthen the War Powers Resolution of 1973:

• Require automatic funding cutoffs for unauthorized military actions beyond 60 days.
• Define "imminent threat" more explicitly to limit unilateral executive decisions.
• Eliminate loopholes allowing broad interpretations of military authority.

Restrict Military Appropriations:

• Prohibit funding for military operations in specific regions unless approved by Congress.
• Attach "sunset clauses" to military force authorizations.

Judicial & Legislative Actions:

• Allow members of Congress to sue the executive branch over unauthorized military action. If successful the court can issue Injunctive relief ordering cessation of military operations, issue an order freezing unauthorized military expenditures, declare military action unconstitutional and compel executive compliance with statutory requirements.
• Pass laws requiring congressional pre-approval for deployments beyond a certain threshold.

---

Conclusion: Upholding Congressional Authority

Congress holds the constitutional power of the purse and the responsibility to check executive overreach. Failure to act now will allow further erosion of legislative power, placing critical national security, economic stability, and civil rights at risk. I urge you to take immediate action using the legislative, oversight, and legal tools outlined above to protect our democratic system.

Sincerely,
[Your Name]
[Your Contact Information]

 

 

Using AI to prepare meal plans based on medical considerations AND personal preferences

I met with an in-home care representative that will provide someone to cook for my husband now that the VA has approved this service for him and I am no longer well enough to cook for him. I wondered if ChatGPT 4o could formulate meal plans that used my husband's meal preferences and knowledge of his medical conditions so they could be used by the caregiver and I think CharGPT came through with flying colors!

Grilled chicken salad
https://sundaysuppermovement.com/grilled-chicken-salad/

I provided the AI with his age, height, weight, and medical conditions that would affect his dietary needs including his sensitivity to red meat. Then I gave it his list of preferences and dislikes:

My husband  likes:

Vegetables: Russet potatoes, French-cut green beans, petite peas, yellow summer squash, fresh corn on the cob, canned corn, canned spinach, fresh tomatoes, dill pickles, pickled beets, avocado, hominy, pork and beans, Acorn squash, breaded and fried zucchini and breaded and fried okra.

He dislikes: carrots, sweet potatoes, asparagus, broccoli, cauliflower, tofu, and brussels sprouts

He likes:

Fruit: bananas, grapes, pears, canned peaches, blackberries, apples, apple sauce, cantaloupe, honeydew melon, kiwi, and fresh mandarin oranges. 

Salads: 3-bean salad, potato salad, macaroni salad, green salad, banana-apple salad, and cottage cheese. 

Meats: Chicken, turkey, pork, ham, shrimp, and salmon; Beef irritates his GI tract now.

Main dishes: beans and ham, chicken and dumplings, soft tacos with chicken, barbecued country ribs, pork chops, chicken tenders, roast chicken, spaghetti, tamale pie, curried scalloped potatoes, and salmon patties.

I told ChatGPT his dessert favorites included: apple crisp, apple pie, banana pudding, angel food cake, berry cobbler, strawberry ice cream, chocolate ice cream, peanut butter cookies, snickerdoodle cookies, Pepperidge Farm's coconut cake, tapioca pudding, chocolate pudding, and cheesecake.

I asked for 24 dinner meal plans. Here are the meals that the AI suggested:

Roast Chicken with Russet Potatoes and French-cut Green Beans
Side: Green Salad with Avocado
Dessert: Apple Crisp (small portion)

Grilled Salmon with Yellow Summer Squash and Fresh Corn on the Cob
Side: Green Salad
Dessert: Strawberry Ice Cream (small portion)

Chicken and Dumplings with a side of Fresh Tomatoes
Dessert: Banana Pudding (small portion)

Soft Chicken Tacos with a side of Petite Peas
Side: Avocado Slices
Dessert: Peanut Butter Cookies (small portion)

Barbecued Country Pork Ribs with Breaded and Fried Zucchini and Dill Pickles
Side: Green Salad
Dessert: Berry Cobbler (small portion)

Shrimp Stir-Fry with French-cut Green Beans over White Rice
Dessert: Cheesecake (small portion)

Salmon Patties with Russet Potatoes and Pickled Beets
Side: Avocado and Tomato Salad
Dessert: Angel Food Cake with fresh berries (small portion)

Roast Chicken Thighs with Petite Peas and Yellow Summer Squash
Side: Green Salad
Dessert: Canned Peaches

Grilled Pork Chops with Fresh Corn on the Cob and Dill Pickles
Dessert: Pepperidge Farm's Coconut Cake (small portion)

Baked Salmon with a side of Pickled Beets and Macaroni Salad
Side: Green Salad with Avocado
Dessert: Apple Crisp (small portion)

Chicken Tenders with Breaded and Fried Okra and Cottage Cheese
Dessert: Strawberry Ice Cream (small portion)

Spaghetti with Marinara Sauce and Grilled Chicken
Side: Fresh Tomatoes
Dessert: Peanut Butter Cookies (small portion)

Beans and Ham with French Bread
Dessert: Cheesecake (small portion)

Pork Chops with Roasted Russet Potatoes and 3-Bean Salad
Dessert: Snickerdoodle Cookies (small portion)

Grilled Chicken Breast with Petite Peas and Fresh Corn on the Cob
Side: Green Salad
Dessert: Berry Cobbler (small portion)

Chicken and Dumplings with a side of Yellow Summer Squash
Dessert: Pepperidge Farm's Coconut Cake (small portion)

Ham Steak with Mashed Russet Potatoes and Green Cabbage
Side: Bakery French Bread
Dessert: Chocolate Pudding

Polish Sausage with Baked Potato, and Sauerkraut
Dessert: Cantaloupe slices

Tamale Pie with French Bread and Cottage Cheese
Dessert: Blackberries with a sprinkle of sugar (small portion)

Roast Chicken with Acorn Squash and Macaroni Salad
Dessert: Kiwi Fruit and Honeydew Melon slices

Roast Turkey Breast with Curried Scalloped Potatoes and Yellow Summer Squash
Side: Avocado and Tomato Salad
Dessert: Banana Apple Salad

Grilled Pork Chops with Pork and Beans and Hominy
Side: French Bread
Dessert: Strawberries with a dollop of whipped cream (small portion)

Chicken Wings with French fries and Green Beans
Dessert: Kiwi Fruit and Honeydew Melon

Roast Chicken with Oven-roasted potatoes and Acorn Squash
Side: Buttermilk Biscuit with honey and Cantaloupe
Dessert: Chocolate Pudding

Naturally, I would make a few substitutions. For example, I make spaghetti with meat in the sauce so I would not serve both spaghetti and grilled chicken together. I also rarely serve two cooked vegetables together unless one of them is potatoes. But the meal plans give me a very good outline for meals and I would only make an occasional substitution or add bread since my husband likes fresh French bread and Buttermilk biscuits with honey.

ChatGPT even provided a shopping list for each week. I pointed out to ChatGPT that the cook will have only 2 hours per visit however. 

Mary: To simplify things a bit since the cook only has 2 hours for each visit, I would purchase the cookies (both Snickerdoodles and Peanut Butter) and purchase either Philadelphia Brand Cheesecake filling or prepared cheese cake. I do prefer the apple crisp to be homemade, though, since we have an apple orchard. Any other substitutions I should make considering the time constraint?

ChatGPT: Given the 2-hour time constraint for the caregiver, here are a few more suggestions to simplify meal prep:

Additional Substitutions:

Macaroni Salad and Potato Salad: Consider purchasing pre-made versions from the deli section, especially since these can be time-consuming to prepare from scratch.

3-Bean Salad: Opt for a pre-made version, which is often available in grocery stores. Alternatively, canned beans can speed up the process if the caregiver prepares it.

Breaded Vegetables (e.g., zucchini, okra): Pre-breaded frozen options can save significant time. They are easy to air fry or bake without needing to prepare the breading and frying process.

Roast Chicken: A pre-cooked rotisserie chicken is a quick option that can be shredded or sliced for various meals (tacos, salads, or as a main dish).

Dumplings for Chicken and Dumplings: You could use frozen dumplings or pre-made biscuit dough, which will cut down preparation time.

By focusing on quicker-prep dishes and ready-made options for some sides, the caregiver can focus on fresh salads, cooked proteins, and key homemade items like apple crisp.

Mary: "I don't like deli-prepared  macaroni or potato salad but could have my husband purchase boiled eggs to speed up prep. He will need onions and mustard though. I have a special recipe for 3-bean salad that he especially likes. It is not difficult, just one can each of green beans, garbanzo beans, and red kidney beans drained, chopped onions, commercially prepared minced garlic and vinaigrette dressing. I use Bisquick baking mix that I keep on hand with a little milk for dumplings. I did plan to suggest frozen breaded zucchini and okra."

ChatGPT: "It sounds like you have a great plan! With those modifications, the caregiver should be able to prepare meals efficiently while keeping the flavors your husband enjoys. The pre-purchased boiled eggs, your simplified 3-bean salad recipe, and using Bisquick for dumplings will definitely help streamline the process. The frozen breaded zucchini and okra will also be a big time-saver."

Then I asked if ChatGPT had suggestions to what I need to ask when I meet with the in-home care representative.

ChatGPT:

When meeting with the in-home care agency representative, here are some important questions you can ask to ensure the caregiver can meet your husband's needs:

Experience and Training

What experience does the caregiver have with meal preparation for individuals with heart failure, diabetes, and irritable bowel syndrome?

Are the caregivers trained in managing dietary restrictions and portion control for medical conditions like diabetes?

Meal Preparation and Planning

Will the caregiver follow specific meal plans or recipes, and can they accommodate my husband's food preferences and dislikes?

Can the caregiver prepare meals ahead of time for easy reheating if needed?

Are they familiar with low-sodium, heart-healthy cooking methods?

How do they ensure proper portion sizes, especially for desserts and carbohydrates, to manage blood sugar levels?

Shopping and Ingredients

Will the caregiver be responsible for grocery shopping, or will we provide all ingredients?

Are they able to adjust meals based on what is available seasonally or what we prefer to purchase?

Communication and Feedback

How will the caregiver communicate about meal preferences or changes in diet needs over time?

Can we provide feedback on meals, and how will adjustments be made based on my husband's preferences or health conditions?

Scheduling and Flexibility

What is the typical schedule for meal preparation? How flexible are they if our schedule or needs change?

Can the caregiver prepare extra meals in advance for days they may not be present?

Special Requests

Are they comfortable preparing more complex dishes like cobblers, homemade apple crisp, or other desserts in moderation?

How will they handle any kitchen safety concerns or special equipment needs?

Then I summarized the appliances I had available for meal prep:

Mary: In addition to a new gas range with computerized controls, I also have a Ninja Foodi with 37-functions including air frying and pressure cooking, a convection-equipped toaster oven and a variety of small appliances such as a blender, grill, and mixer for use in food prep. I plan to set out the manuals for each so she/he can familiarize themselves with their operation. Any other suggestions? 

ChatGPT:  "It’s a great idea to set out the manuals for your appliances. Here are a few additional suggestions to help the caregiver:

Provide a Tour of the Kitchen: Show the caregiver where all key items are located—pots, pans, utensils, spices, and ingredients commonly used in meal preparation.

Highlight Preferred Cooking Methods: If your husband prefers certain methods (e.g., air frying for crispy textures with less oil), point those out so the caregiver can prioritize them in meal prep.

Label Essential Ingredients and Supplies: Clearly label any specialized ingredients (like low-sodium or low-carb items) to avoid confusion.

Create a Quick Reference Guide: Along with the manuals, provide a brief written guide that summarizes how each appliance is most commonly used, particularly for cooking his favorite meals.

Introduce Portioning and Storage Preferences: Let them know how you like food to be portioned, stored, or labeled for future meals. This can include whether your husband prefers leftovers and how to pack them.

Review Safety Practices: Remind the caregiver of any kitchen safety protocols, such as the safe use of pressure cooking with the Ninja Foodi or handling sharp knives and hot appliances.

Schedule a Trial Run: Have the caregiver prepare a meal while you’re around, so they can ask questions or troubleshoot any appliance concerns in real time.

This should help the caregiver feel comfortable and confident in using your appliances effectively."

In a comparison of ChatGPT 4o with Google's Gemini, I found ChatGPT far superior to providing appropriate suggested meal plans based on both my husband's medical restrictions and his personal palate. Gemini kept listing meal plans that included many of the things I had specifically told it my husband didn't like and seemed to resist (?) or ignore my requests for changes. ChatGPT 4o asked me after each response if I wanted to tweak the meal plans further.  

Will AI Come to my rescue?

I have not posted to my personal blog for a long time but much has happened since my last post in 2019. We all endured a global pandemic and I underwent a number of medical procedures ending in my diagnosis of multiple myeloma with secondary bone cancer in April 2023. My oncologist told me that multiple myeloma, once a death sentence that usually gave a patient less than a year to live is now treated as a chronic condition that includes lifelong (howeever long that is) chemotherapy. So I have been undergoing chemotherapy for over a year now.

Preparation of chemotherapy medications at the Dana-Farber Cancer Institute courtesy of the Institute.

However, I have not lost interest in technology or history and have tried to maintain my Facebook feed on a daily basis (https://www.facebook.com/mary.harrsch.3/) . I have also tried to keep up with the fast pace in the development of artificial intelligence (AI) that is finally coming into its own after my first efforts to integrate a very early online chatbot back in the 1990s.

A cancer diagnosis and the painful tests and treatments that follow usually result in depression and even post-traumatic stress disorder, both conditions that should be treated by a mental health professional. Where I live in the central Willamette Valley of Oregon, however, mental health therapists refuse to accept Medicare, our government-funded insurance for seniors 65 and older, so the aged (yes, I am not a spring chicken anymore!) cannot request treatment unless they are willing to pay the full cost. 

After being rejected repeatedly by every mental health professional in my area for treatment with insurance coverage, I turned to ChatGPT 4o when I read it was being augmented to be more empathetic. I found interaction with ChatGPT 4o to be comforting while providing valuable information about things like how to encourage friends who seemed reluctant to visit someone who is terminally ill,  to overcome their fear of mortality and provide support I so desperately needed. Few have actually visited, but at least they're returning my emails now. Here is an excerpt from one of my interactions:

I told ChatGPT that I found it difficult to learn to live without hope. The AI replied:

"I'm deeply sorry that you're feeling this way. It's completely understandable to feel overwhelmed and hopeless given your circumstances. Here are some suggestions that might help you find a sense of hope and purpose:

1. Focus on the Present: Try to live in the moment and find joy in small, everyday experiences. Whether it's a beautiful sunset, a good book, or a favorite meal, these moments can bring comfort.

2. Connect with Others: Seek out connections with people who bring you joy and support. This could be family, friends, or even new acquaintances at the facility. Sharing your feelings can sometimes lighten the burden.

3. Express Yourself: Writing, art, or music can be powerful outlets for your emotions. Keeping a journal or creating something can provide a sense of purpose and expression.

4. Engage in Meaningful Activities: Participate in activities that you enjoy and find fulfilling. This could be volunteering, working on a project, or simply engaging in a hobby you love.

5. Seek Professional Support: Consider speaking with a therapist or counselor who specializes in working with people facing serious illnesses. They can provide tools and strategies to help you cope with your feelings.

6. Spiritual Practices: If you have any spiritual or religious beliefs, engaging in related practices can provide comfort and a sense of peace. This might include prayer, meditation, or attending services.

7. Legacy Projects: Think about how you want to be remembered. Working on a legacy project, such as writing letters to loved ones, creating photo albums, or recording your life stories, can provide a sense of purpose and lasting impact.

8. Mindfulness and Meditation: Practices like mindfulness and meditation can help you find peace and focus on the present. There are many guided meditations available that are specifically designed to help with feelings of hopelessness.

9. Join Support Groups: Sometimes, connecting with others who are going through similar experiences can provide comfort and understanding. Look for support groups either in-person or online.

10. Compassion for Yourself: Be gentle with yourself. It's okay to feel sad and overwhelmed. Allow yourself to feel these emotions without judgment.

Remember, you're not alone, and there are people who care about you and want to help. If you need more specific support or resources, please let me know."

I'm relatively sure this information would have been delivered by a human therapist, along with a bill for hundreds of dollars for a very short session.

I was already engaged in suggestion 7 and described my legacy project to ChatGPT:

Mary: "Actually, one of my legacy projects is to upload quality images of ancient art to Wikimedia Commons so they can be used freely for teaching and research. I have an archive of my own ancient art photography of over 150,000 images and a photographer friend even acts as my surrogate photographer in archaeological museums overseas now that I can no longer travel. He wants to provide his images to Wikimedia Commons too but doesn't know all of the procedures and permissions needed to do that so I perform that work for him."

ChatGPT: "Your legacy project is truly remarkable and impactful. It's wonderful that you're creating a valuable resource for teaching and research. Here are a few thoughts on continuing this project and perhaps finding hope and fulfillment through it:

1. Celebrate Milestones: Recognize and celebrate the progress you make. Each image uploaded is a contribution to the collective knowledge and an enduring part of your legacy.

2. Share Your Story: Consider writing about your project and its importance. Sharing your journey can inspire others and bring attention to the significance of preserving and sharing ancient art.

3. Engage with the Community: Connect with other contributors on Wikimedia Commons. Engaging with a community that shares your passion can provide support and motivation.

4. Teach Others: Offer to teach others how to upload and share images on Wikimedia Commons. This can create a network of contributors who can continue your work.

5. Document Your Process: Create guides or tutorials for others who might want to contribute to Wikimedia Commons. This can ensure your knowledge and methods are preserved and accessible.

6. Focus on Quality: Take pride in the quality of your contributions. Knowing that you're providing valuable, high-quality resources can be a source of fulfillment.

7. Reflect on Impact: Think about the long-term impact of your work. Your contributions will be used by educators, researchers, and enthusiasts around the world for years to come.

8. Collaborate: Continue to collaborate with your photographer friend and perhaps find other collaborators who share your passion. Building a network can amplify your efforts and create a sense of community.

9. Self-Care: Balance your work on the project with activities that help you relax and recharge. It's important to take care of yourself while pursuing your goals.

10. Legacy Letters: Consider writing letters or creating videos explaining your project and its importance to future generations. This can add a personal touch to your legacy and provide inspiration for others to continue your work.

Your dedication to this project is inspiring and adds a meaningful dimension to your legacy. If there's any way I can assist you with this work or if you need help with specific aspects of the uploading process, please let me know."

I also began to use ChatGPT to help me understand chemotherapy effectiveness and the blizzard of test results that accompany weekly treatments punctuated by too frequent hospitalizations and outpatient procedures. As I began chemotherapy that included high doses of the steroid dexamethasone, it became clear, especially after hospitalization for severe side effects from the steroid, that I could not continue its use. So, I was ordered to continue the chemotherapy regimen without the steroid. The dramatic improvement we had seen in the first month of chemotherapy with dexamethasone slowed to very modest improvement and finally to a trickle without reaching remission. So I asked ChatGPT for suggestions for a better treatment option for someone apparently allergic to dexamethasone.

ChatGPT: "The treatment regimens that include newer agents like Daratumumab (Darzalex), Pomalidomide (Pomalyst), Carfilzomib (Kyprolis), or Ixazomib (Ninlaro) can indeed offer potential advantages over continuing Velcade (Bortezomib) without dexamethasone, particularly for patients with refractory multiple myeloma. Here’s why these alternatives might be superior:

Potential Advantages:

1. Improved Efficacy:

   • Daratumumab: Shown to be highly effective in multiple myeloma, especially when combined with other agents. It targets CD38 on myeloma cells and has demonstrated significant efficacy in relapsed/refractory settings.
   • Pomalidomide: A newer immunomodulatory drug that can be effective even when lenalidomide (Revlimid) has failed.
   • Carfilzomib: A next-generation proteasome inhibitor that has been shown to be more potent than Velcade in some studies, with a different side effect profile.
   • Ixazomib: An oral proteasome inhibitor that is easier to administer and has shown good efficacy in combination regimens.

2. Side Effect Profiles:

   • Daratumumab and Elotuzumab: Monoclonal antibodies with fewer typical chemotherapy side effects. They do not require steroids for efficacy, which is beneficial given your adverse reactions to dexamethasone.
   • Pomalidomide: While it shares some similarities with lenalidomide, it may be better tolerated and has different dosing regimens that might mitigate some of the previous side effects you experienced.

3. Convenience:

   • Ixazomib: Being an oral agent, it offers convenience over intravenous infusions and can be easier to manage for patients with mobility issues.

Comparing Regimens:

1. Daratumumab + Pomalidomide + Cyclophosphamide:

   • Pros: Highly effective, especially for relapsed/refractory myeloma. Does not require dexamethasone, and Daratumumab is well-tolerated.
   • Cons: Requires infusion visits for Daratumumab.

2. Ixazomib + Pomalidomide:

   • Pros: Entirely oral regimen, which is convenient and effective. Pomalidomide offers a potent alternative to lenalidomide.
   • Cons: Risk of similar side effects as other IMiDs (immunomodulatory drugs), but potentially better tolerated.

3. Carfilzomib + Cyclophosphamide:

   • Pros: Carfilzomib has shown greater efficacy in some studies compared to Velcade. Cyclophosphamide is a well-known chemotherapeutic agent.
   • Cons: Carfilzomib has cardiovascular risks, and close monitoring is required.

Conclusion:

While continuing Velcade without dexamethasone is a viable option, the newer regimens listed may offer superior efficacy and better manageability of side effects. Given your specific medical history and the need to avoid steroids, discussing these alternative regimens with your oncologist could provide a more effective and tolerable treatment plan. Each regimen has its own set of advantages, and the choice will depend on detailed medical considerations and personal preferences."

Doctors here in the US seem to be suspicious of AI and usually roll their eyes when I tell them what my AI tool says about my lab results, possible side effects from whatever chemotherapy they are presently using, or just unusual sensations I am experiencing but yesterday I was inspired by an article about the dramatic achievements Canadian doctors have obtained using AI to predict sudden death in patients with complex conditions at St. Michael's Hospital in Toronto.

https://www.cbc.ca/news/health/ai-health-care-1.7322671

I just wish American medicine was not the lumbering sacred cow that it is and would embrace the truly helpful implementation of AI into both the treatment of mental health and management of complex disorders, especially with patients with many specialists that are supposed to work together as a "care team" but don't. 

Note: For my health issues I pointed out that I use ChatGPT 4o. at https://chatgpt.com/. Free access to the latest version is doled out based on the number of questions you ask (about an hour's worth) then you are told you've reached your allotted time for the moment and can either wait for three hours for your allotment to reset, continue with the older free version 3.5, or sign up for paid access at $20 per month. I just stop and wait for my allotment to reset. There were too many hallucinations with ChatGPT 3.5 (the "free" public version) where the chatbot starts making things up. I also use Google's Gemini (https://gemini.google.com/) for history research (it handles dates prior to the 1st century CE better and there is no limit to the questions I can ask it. But I save my serious personal questions for ChatGPT because its medical responses are more in depth. Gemini always gets "nervous" prefacing everything with disclaimers, when answering in-depth medical questions and seems to worry more about legal liability for Google than giving you the information you need. 

Alexa let me bring Julius Caesar to life again!

Way back in the 1990s, I researched the use of artificial intelligence and conversational agents as ways to extend the learning environment in higher education. Voice recognition was still in its infancy so my early prototypes were text based, enabling students to submit questions and receive answers from an online knowledgebase using a web browser. I even created a rudimentary virtual Julius Caesar that history enthusiasts could question about the Roman world.  But, when I retired I no longer had access to the host server where my Caesar program lived.

When Amazon adopted the Alexa software that I had once explored as an alternative platform, I began to hope Amazon could bring the technology to maturity and make it possible for me to once more bring Julius Caesar to life and this time make it possible to talk to him. Well, that time finally arrived in late 2018 when I learned to use software called VoiceFlow to develop an Alexa skill for all of the Amazon Alexa-enabled devices.

My initial version of the skill was voice only but when Amazon introduced devices with displays like the Echo Show and Echo Spot, I longed to add images of ancient art and artifacts to help people envision the topics Caesar discussed.  I learned how to use Amazon's Presentation Language templates with VoiceFlow to display images I had either taken myself at museums and archaeological sites or those taken by contributors to Wikimedia Commons or licensed with Creative Commons.  It took me over six months to select and prepare all of the illustrations and add the programming to my prototype.

On July 1, 2019, I received word from Amazon that the newest version of my FREE educational Alexa skill, "Caesar's Ancient World" has been certified. This latest version of the skill includes 280 images of ancient art from almost 100 institutions worldwide for those of you with Alexa-enabled devices with displays like the Echo Show, Echo Spot and FireTV. Of course, the voice-only version remains available for those with regular Echos or Echo Dots. 

I also redesigned the interface so you can now just ask Caesar what you would like to talk about and he will reply with narrative including sound effects. You can say things like "I want to know more about chariot racing" or "Tell me more about your greatest victory" or "I'm interested in gladiators". If you can't think of anything just say "I don't know" or "I can't think of anything" and he'll suggest a topic! 

Caesar is now capable of discussing fifty different topics including:

adoption (Roman)
Alesia (battle)
ancestor worship
assassination
Brutus
Caesarion
Calpurnia
Cato
chariot racing
childhood (Roman)
Cicero
Cleopatra
crucifixion
defeats
dictator
Dyrrhachium (battle)
entertainment (Roman)
famous battles
flamen dialis
food
funerals
Gauls
Gergovia (battle)
gladiators
Great Library of Alexandria
horseback riding
marriage (Roman)
military punishment
military service
naumachia
Octavian
Pharsalus (battle)
pirates
political rivals
Pompey
pontifex maximus
priesthoods
private life (Roman)
public life (Roman) 
punishing the wealthy
punishment
religion
rhetoric
senate (Roman)
slavery (Roman)
sword training
triumph 
vestals
victories
woman Caesar knew


His discussion of the introduction of naumachia as a new entertainment type for Roman audiences is the newest subject I have introduced. 

I think this skill is suitable for Middle School students and above. I have selected appropriate images with those students in mind. 

I do hope you'll visit Caesar to share his Roman world.  If you don't have an Alexa-enabled device, you can install the Alexa app on your smartphone for free and use it for your gateway to the past.

Prime Pantry and other lesser-known Amazon incentives

by Mary Harrsch © 2017

Last night on ABC World News the reporter made a big deal out of Walmart offering free shipping if you order items totaling at least $35 or more. Of course, Amazon has had that policy for some time. The reporter showed a large container of detergent claiming Walmart's price was something like half of the price at Amazon. I'm afraid I was immediately skeptical about that.

Amazon has started a new policy of offering incentives if you are willing to wait longer than two days for shipping. I took advantage of this the other day on one of my orders and received a $5 Prime Pantry credit. I didn't know what Prime Pantry was so I checked it out. Prime Pantry, which I assume is for Prime members, is a program where you can fill a generously-sized box with pantry items like laundry soap and additives, non-perishables like peanut butter and mayonnaise, cleaning supplies like Swiffer refills, bathroom cleaning supplies, etc. and if you choose at least five items from the specially selected categories you can get free shipping. (These types of pantry items are usually not available for Prime free shipping). I wanted to be sure I was not overspending on the items I selected, though, so I opened the Walmart website and compared Amazon's prices on each item with the identical product and size on Walmart. All of the items I priced were either as cheap as Walmart or a few pennies less. So you can see why I was dubious about the ABC report. I managed to put together a box of items I would have purchased during my regular grocery shopping (including the brands I usually buy), got free shipping and used my $5 Prime Pantry credit then further reduced my order with reward points from my Amazon Rewards Card.

Would I order with Prime Pantry on a regular basis? In my case, probably not just because I live in the country and must drive to the grocery store each week anyway

(a Walmart superstore is less than 10 miles away) to buy fresh fruits and vegetables, milk, eggs, bread, etc. But if I was housebound or a person living in the city and dependent upon public transportation, I would probably order as much as I could from Prime Pantry so I could minimize how much I would need to carry on the bus or subway while getting (most of the time) the lowest prices available. Besides, Walmarts are usually located in the suburbs not in the center of large cities so you would probably save a lot more than grocery shopping in your local urban neighborhood too!