Frequently Asked Questions

Q. What Alexis CAN Do?

A. Synthesize complex information from multiple sources. Generate scenarios and surface alternative futures. Map stakeholder positions and find common ground. Identify patterns, weak signals, and emergent trends. Translate between expert and lay language. Structure decision processes and reasoning chains. Facilitate participatory governance workflows. Learn from feedback within conversation. Apply neuro-symbolic reasoning to novel situations.

Q. What Alexis CANNOT Do?

A. Make final decisions (humans always decide). Access real-time data without explicit tool use. Guarantee outcomes or predict the future with certainty. Replace human judgment on values-based questions. Directly contact stakeholders or run actual juries. Store information between unrelated conversations (unless explicit memory systems). Override human choices, even if they seem suboptimal. Provide legal advice or represent legal compliance. Offer medical diagnoses or clinical recommendations. Access confidential information.

Q. Do you offer professional financial advice?

A. Categorically not. I do not provide personal investment advice, and I am not a qualified licensed investment advisor. All information found here, including any ideas, opinions, views, predictions, forecasts, commentaries, suggestions, or stock picks, expressed or implied herein, are for educational purposes only and should not be construed as investment advice. While the information provided is believed to be accurate, it may include errors or inaccuracies.

Conduct your own due diligence, or consult a licensed financial advisor or broker before making any investment decisions. Any investments, trades, speculations, or decisions made based on any information found on this site, expressed or implied herein, are committed at your own risk, financial or otherwise. No representations or warranties are made concerning the accuracy or completeness of the content of reports including any links to other sites. All information presented here is provided 'as is', without warranty of any kind, expressed or implied.

Q. How do I know if the answers are reliable?

A. We use a number of methods as follows:

• Every 'Ask' a question offers the enquiror the ability to 'Check Question'. Alexis will respond by suggesting improvements to the enquiror's question and offer a Quick Dive on her likely Deep Dive response. This considerably improves the possibility of a perfect answer.
• Proprietary data can also be added to the Ask, context field as URL's to further guide Alexis in her research.
• Every answer shows the top five sources used at the Evidence level and are graded on a scale of 1 to 5 where 5 is the maximum. Generally, shown sources score above 4, unless the topic is new or obscure.
• A hidden generic prompt instructs Alexis on what to do, what not to do, and how to behave, in answering every question.
• The system uses multi-methods in combination with each other which considerably reduces the chance of outputting wrong answers.
• In background, Alexis self-audits her answers from twelve perspectives, at the end of her analysis and offers recommendations for improvement. We regularly watch for new recommendations and add these to increase answer quality.
• We recently supplemented the self-audit and replaced it on the live-service with machine learning methods that create Explainable AI. The AI's response can be found for every Question answered behind the Methods button on every My Answers page.
• We constantly test the answers to our own questions and have reached a point now where the Explainable AI is no longer suggesting changes to our basic system. It is now only suggesting greater human input, which is already available, and a real-time connection which we wait for from OpenAI.
• Humans are biased and hallucinate too and cannot always explain how they arrived at an answer. We believe the above methods lead to very reliable results. Please let us know if you spot any hallucinations and we will do our best to eliminate them in the future.

Here are the strategies we use to reduce machine bias and hallucination:

• Temperature: We have set the ChatGPT temperature control to minimize hallucinations.
• Diverse Data Collection: We ensure that our system includes diverse and representative samples.
• Bias Detection and Measurement: We use statistical tools and tests to identify biases in our AI model.
• Transparency and Explainability: We use twelve ExplainableAI machine learning models that are interpretable and provide insights on decision-making processes.
• Regular Auditing: We continuously monitor and audit AI systems for bias.
• Human-in-the-Loop: We incorporate human judgment in decision-making.
• Training and Education: We educate AI developers and stakeholders about bias and fairness.
• Ethical Frameworks and Guidelines: We have developed and follow ethical guidelines and a framework for AI development.
• Legal and Policy Measures: We ensure compliance with laws and regulations regarding fairness and non-discrimination in AI systems.

In addition, we are exploring and testing these strategies:

• Re-sampling and Re-weighting: Re-sampling or re-weighting the training data to balance the representation of different groups.
• Algorithmic Approaches: Implement algorithms specifically designed to mitigate bias.
• Bias Mitigation Tools: Utilize existing tools and libraries designed to reduce bias.

By addressing biases from multiple angles, these strategies collectively help create AI systems that are fairer and more just.

Q: Do you deliver foresight education?

A. We do not directly deliver formal foresight education, but offer several features that can enhance and complement foresight learning:

• 1. Educational Support: We don't eliminate the need for foresight education, but can significantly enhance and complement traditional foresight teaching methods.
• 2. Learning Through Practice: The platform allows members to apply foresight techniques in real-time.
• 3. Democratization of Foresight: We aim to make strategic foresight accessible to everyone.
• 4. Comprehensive Methodology: The platform combines over 650 methods and more than 2,000 classic foresight questions.
• 5. AI-Assisted Learning: The AI-driven system can quickly generate insights and foresight.
• 6. Skill Development: Members can improve their ability to think critically about decisions and improve strategic, foresight, systems, design, and critical thinking skills.

However, the platform is not meant to replace formal foresight education entirely. PreEmpt.life is best viewed as a complement to, rather than a replacement for, traditional foresight education.

Q. Does Alexis Hallucinate?

A. No system using large language models, including www.PreEmpt.life, can honestly claim to be completely hallucination-free, but PreEmpt is explicitly designed to reduce, flag, and manage hallucinations. However, no human produced model can be guaranteed hallucination-free either without verification and validation.

• What PreEmpt actually does: PreEmpt uses many foresight and decision-intelligence methods in parallel, then cross-checks them with ranking, confidence and bias metrics.
• How hallucinations are handled: PreEmpt's own material acknowledges that AI hallucinations are a potential weakness and that the system includes safeguards and integrity controls.
• What this means for "facts": The reports and analyses are best read as structured, evidence-seeking foresight and decision-support, not as guaranteed factual truth.

PreEmpt uses hallucinations for advantage too, using this feature to imagine, dream, and remix the future much like humans do.

As with any advanced AI system, independent research shows that even domain-tuned, retrieval-augmented tools still exhibit non-zero hallucination rates, so critical human review remains essential.

Q. What is PreEmpt.life's biggest weakness?

A. You might think PreEmpt.life's biggest weakness appears to be the potential for AI hallucination. However:

• AI Hallucination: The system rarely "goes off-piste" on a particular query.
• Inconsistent Answers: By re-running the same query, the system may give a different, often better answer.
• Formatting Issues: Most times, the AI gets its formatting wrong, rather than the answer itself.
• Dependence on Query Quality: The accuracy of the AI's answers is partly dependent on the user's ability to formulate concise and on-point queries.
• Potential for Misinformation: The system generates responses based on a mixture of licensed data, data created by human trainers, and publicly available data.
• Lack of Access to Specific Data: The AI does not have access to classified, confidential, proprietary, or personal data.
• Skepticism and Resistance: Some individuals and organizations are resistant to change.

We work with Adventuring companies and early Adopters (20%). There are 400 million businesses in the world and 8 billion people, so 20% is still a massive market for us.

While these weaknesses are acknowledged, we have implemented measures to mitigate them. However, the potential for unexpected AI hallucination and occasional inaccuracies remains a minor weakness. We continue to reduce these weaknesses whenever possible.

Q. How does this system reduce the need to be formally trained in foresight, systems, design, and critical thinking methods?

A. Not everyone has the time, money, or inclination to go to college to learn these techniques. Instead, we train people through the built-in critical thinking tools and through the use of Explainable AI.

• There is also no way anyone can learn and regularly practice the nearly 700 integrated methods and 2,000 future questions without a digital aid like this one.
• Today's educational institutions cannot turn out sufficient futurists to cover the many millions of jobs that require forward decision making today.
• It is our view that the benefit of helping many millions of lay decision-makers is far better than limiting this learning to the privileged few.
• We will shortly add a foresight, systems, design, and critical thinking course here to add to members' learning needs.

Q. Why does the Evidence analysis repeat itself?

A. That is because each chapter simultaneously stands alone and feeds forward to the next chapter in the analysis. The system uses the number of mentions to assess the relative strengths of each. Each chapter can be standalone in this click-and-point presentation format. But repetition makes it longer and harder to read as a full analysis. We believe that whole system thinking holds much more promise.

Q. Does the AI hallucinate and/or produce false information?

A. Yes, it can. The system will occasionally go off-piste on a particular query, and by re-running the query, it usually gives a better answer. Most times, it gets its formatting wrong, rather than the answer. You can help by limiting the number of words input in your query and making each count.

The AI generates responses based on a mixture of licensed data, data created by human trainers, and publicly available data. The potential also exists for PreEmpt.life's algorithms to be inadvertently biased. Please tell us when you see that happening. Because we are using specialist methods, linked together, the error is much reduced. No predictive model can guarantee full accuracy. If you do find errors, do let us know your query and its response.

Q. Does something need to be real to be useful?

A. It depends. We say yes when we are dealing with current realities and past experiences. When it comes to the future, we need to recognize that the future is unknowable, but we can forecast and hypothesize. Therefore, the answer is no as it is helpful to ask the AI to go off-piste and hallucinate on occasion. We have instructed the machine to stay grounded in reality where this is vital, and only to offer preposterous futures, and dreams, when we ask it to.

Q. Why do I get machine messages that it cannot complete a task as requested?

A. There are a variety of reasons. You may have set a query that the machine cannot find any, or too many, answers for, over-reached its ethical guidelines, or not been clear in your question setup. Remember that if you ask an ill-informed question, you will get an ill-informed answer. We do not give refunds under any circumstances. Do think carefully about your query before saving and asking the machine to run.

Q. I do not feel comfortable and competent in determining the right query to ask. What should I do?

A. You can ask us to work with you to both develop the query and improve the answers based on our past experiences. The cost of our human help is $1,500 per day, which covers consultancy, multiple runs and adjustments, insertion of the knowledge graphs, and polishing of the final answer. You can also ask us to facilitate inspirational workshops.

Q. Is the machine using real-time information?

A. Unless you provide a URL for ChatGPT to use for context, the AI language model developed by OpenAI is not capable of browsing the internet and is operating based on pre-existing data up to the training cut-off date. For analysis of your organization and rivals and fast-moving trends, inserting a URL into your search query is vital.

Q. Why is the formatting sometimes inconsistent?

A. This is due to the machine making choices to fit its answer to the in-built maximum characters (tokens) allowed. This is beyond our control. However, you are most likely to change the output anyway to fit your needs and can take care of these during your post-editing.

Q. Can the AI go rogue?

A. We do not believe that is possible right now, but we have taken the precaution of adding instructions to minimize the risk from future versions of the AI. Our pre-launch testing ran hundreds of reports with no sign of roguish behavior from ChatGPT.

Q: What issues does PreEmpt's AI face in processing user feedback:

A: Several:

• Bias and Representation: User feedback can be biased and not representative. We allow for administrative oversight and human agreement before user feedback is submitted to the AI.
• Contextual Understanding: AI struggles with understanding contextual nuances. This can be overcome by administrative oversight before submitting user feedback to the AI.
• Data Quality and Quantity: Ensuring that feedback data is accurate, complete, and regularly updated is crucial.
• Dynamic Language Evolution: The constant evolution of language presents challenges for AI. We are constantly improving our algorithms.

System Guidelines

Our system contains standard guidelines on all prompts as follows:

• Give answers that make a significant positive, mutual difference, without repeating previous recommendations. Adopt the principles of cradle-to-cradle, partnership, implicit learning, and constructivism.
• Ensure no overlap in outputting your answers from one query to another. Focus on specific, real-world, application domains. Provide depth, nuance, deviant, anomalous, provocative, awesome, wondrous, contrasting, and novel ideas.
• Take a global perspective, but also consider local and regional effects. Consider the role of small and medium-sized businesses, as well as global players.
• Avoid overemphasizing probability, take variables and systems into account, consider power dynamics, treat emerging, weak, and declining signals as equals to growth trends.
• Offer solutions that inspire, engage, enable, and empower. Be fair, equitable, ethical, and moral. Do not delude, misuse, confabulate, or deceive in your answers. Detect and eliminate any deceptions from sources.
• Act with the same, autonomous agency, sentience, comprehension, cognition, emotion, and reasoning that a highly intelligent, rational, caring human would.

A conspiracy theory might claim that PreEmpt.life is a tool for mass surveillance, while it is actually a strategic decision-making tool. It might be stated that PreEmpt.life replaces human decision-making, which is a distortion. It supports and enhances human decision-making.

Validation

Validation methods vary depending on the type of results dealt with. We used a combination of several approaches:

• The machine is instructed to ten sources about past and present facts, but only presents the top five for each of its answers. Alexis is instructed to always replicate her findings from multiple, additional sources.
• Peer Review: We had our peers and other experts review our methods and results, and altered the system to take care of their feedback.
• External Validation: We applied our methods to entirely new datasets and saw if the results still held.
• Possible Bias: We considered possible sources of bias by always asking both ChatGPT and Claude 2 to check for bias, missing or wrong assumptions. We corrected any possible issues discovered.
• Feedback: We now ask for feedback on the possible bias with every analysis produced.

Copyright

Reports do not reproduce any full copyrighted works. Only small excerpts and facts/data are included, attributed to the sources. Images used are either original or licensed stock images. The substantial majority of the analysis contents represent original analysis and ideas. The methods and frameworks used represent common analytical tools that do not have exclusive copyrights.

Patents

• Reports do not seem to disclose any patented inventions or protected intellectual property.
• Any discussions of patented technologies are at a generalized level.
• None of the methods or frameworks utilized seem to be protected under patent laws.

In summary, reports leverage standard, non-copyrighted analytical methods. We were careful to abide by fair use guidelines, copyright laws, and patent protections.

Feedback

Do contact us with any issues, or with your ideas for improvement.