{"id":19190,"date":"2025-01-25T14:00:11","date_gmt":"2025-01-25T14:00:11","guid":{"rendered":"\/?p=19190"},"modified":"2025-01-25T14:26:12","modified_gmt":"2025-01-25T14:26:12","slug":"why-do-good-projects-fail-and-good-things-take-decades-when-the-issues-can-be-clear-in-hours","status":"publish","type":"post","link":"\/?p=19190","title":{"rendered":"Why do good projects fail and good things take decades, when the issues can be clear in hours?"},"content":{"rendered":"
\n
\n
Did you try this? Fast accurate calculations are helpful, but the space of possible things to investigate and try is still large. For a variety of reasons, “small parts can be stronger”, “simple shapes take less time”, “many bonds can make a mess” – it makes sense to assemble and use molecular parts, rather than proteins. Not all the way down to polyatomic atomic chains, or to coated inorganic substrates, but to addressible and traceable parts. Can you replace organs? Can you replace membranes or organelles? Can you replace blood or fluids or oxygen carriers? The list of attractive, possibly useful, and possibly lucrative or helpful things on the Internet can still take a long time.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
What is it that you want to make? What do you want to do? Who do you want to help? Or do you simply want money fame and glory? It is hard to do “everything” even when AIs and tools now make much that was impossible last month, possible now. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Do not make up a list of targets where you have biases and guesses what is valuable, and what way to go. Without bias, pick something and solve it. For others, not for yourself. That is closely related to why outsiders see more clearly – their passions, lusts and disgusts are not activated below their level of awareness.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
That way you can learn the methods of “assembly and fabrication design management” and help all groups, not just a few.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
I first heard about ab initio methods in high school in 1963. It was new and attractive and “Let the computers do it” sounded nice. But decades and decades later, it is still a myth and a hope, and real solutions end up being “next year”. Because no one is addressing the very real human dynamics of motivations, rules, priorities, and just plain human greed, laziness, passions and just plain human mistakes. If you make a company or group – unless the whole, its whole industry and the context are just right most things fail for external reasons.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
When the US or Japan or China or Russia or India or Africa or where-ever says “we can concentrate our efforts and do X”, they usually fail because the problem cannot be solved unless everyone is involved fairly and no human group ever sets the rules so they cannot be games by a few. No one, no group, never in human history. That is what I have tried to explore, understand and “try to find solutions for” – “Why do all global and systemic projects take decades to map out, design, and complete? when the computers now can do it in a few hours or weeks?<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
===============<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n
Full converstaion in more readable form at <\/span>https:\/\/chatgpt.com\/share\/6794f29b-8abc-800b-a7e8-94b6e721307a<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Be aware that by default OpenAI always scabbed “good” text from the free Internet and that is highly biased towards “hire we are experts or buy this” . So “form a human committee is almost always a knee-jerks solution offered. These long outputs and their prompts can be combined and “compiled, analyzed and resolved as a whole” to get a more complete and comprehensive maps of what to do. It has to be lossless, and completely traceable or it is a waste of time.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
==========<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Here is OpenAI ChatGTP Plus running ChatGPT o1 says:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Thought about human dynamics in problem-solving for 11 seconds<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Below is a discussion that tries to tease apart why large, complex, global projects often take decades\u2014despite the fact that modern computation can theoretically solve the \u201ctechnical\u201d component in a fraction of the time. It also touches on the interplay of motivations, rules, resources, and the countless human factors that slow (or stall) global coordination.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
1. The Technical vs. Socio-Technical Divide<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Purely Technical Complexity Even if we imagine unlimited computational resources, perfect ab initio methods, and full AI-assisted discovery, there\u2019s still an inescapable layer of human involvement in every real-world project. For example:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Accuracy of ab_initio_methods ~ f(high_fidelity_models, computing_power, time, quality_of_data)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Yes, modern computers can run quantum chemistry, protein folding, fluid dynamics, or advanced optimization. Yet \u201chanding over\u201d a design to an algorithm is just one piece. Even once the simulation yields a perfect structure (a new material, a molecule, a drug), there are multi-year steps to physically validate that result, optimize production, and gain acceptance in real markets.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Socio-Technical Complexity Massive global or systemic projects (e.g., developing a new organ replacement technique, deploying a planet-wide climate technology, redesigning entire supply chains) involve people, markets, policy, culture, law, and international relations. If the core technical solution is only 10% of the effort, the other 90% is:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Socio_technical_challenges = f(political_will, regulations, funding, cultural_acceptance, legal_infrastructure, workforce_skills, supply_chain_capacity, trust)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
No supercomputer can \u201csolve\u201d trust, conflicting priorities, or cultural differences. And if you cannot solve those, the project doesn\u2019t reach scale or momentum.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
2. Conflicting Incentives, Rules, and Alignment Problems<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Individual vs. Global Benefit To solve something \u201cfor the world,\u201d many vested interests need to come together. But each participant\u2014from private companies seeking profit, to governments seeking political advantage\u2014inevitably has its own constraints:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Incentives(each_actor) = f(financial_rewards, political_capital, risk_aversion, internal_competition, national_security)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Global-level projects stall when each major actor can veto, delay, or divert resources to preserve its own advantage\u2014or avoid risk.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Fairness and Perceived Fairness Large-scale efforts unravel when any group feels exploited or excluded. Designing systems that are mutually beneficial, or at least perceived that way, often requires an immense amount of negotiation:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Global_cooperation = f(transparency, shared_governance, distribution_of_benefits, checks_against_exploitation)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If one party believes the rules are \u201cgameable,\u201d or that others hold disproportionate power, deals fall apart\u2014regardless of how good the technical plan is.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
3. The \u201cHuman\u201d in Human-Driven Projects<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Motivations, Emotions, and Biases Rational analysis can fail if it runs counter to ingrained passions, cultural identities, or entrenched biases. People may reject, delay, or sabotage solutions that threaten their sense of identity, moral framework, or financial stake.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Organizational Inertia and Fear of Failure Even big governments with massive budgets are bound by processes, laws, risk aversion, and the fear of being blamed for mistakes. That inertia can delay or derail sweeping changes.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Short-Term vs. Long-Term Vision Politics and markets often favor short-term gains over long-term benefits. Doing \u201cthe big right thing\u201d that pays off in 10\u201320 years can be politically or financially unpopular if it hurts immediate profits or election cycles.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
4. Why Projects Still Take Decades<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Iterative Testing and ValidationEven the most elegant design must pass safety tests, regulatory hurdles, or real-world pilot phases. In medicine, new organ substitutes or advanced therapies can take a decade to gain regulatory approval. Pilot projects, scale-up, and large trials are measured in years, not months. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Global Scale and Coordination<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
The time it takes to align laws, standards, supply chains, and cross-border collaborations stretches on. Each layer of \u201cglobal synergy\u201d can add years. If the project depends on uniform adoption (e.g., every country adopting a policy or technology), the slowest or most resistant stakeholder controls the overall timeline. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Economic Uncertainty and Funding<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Funding can evaporate if investors or governments see a dip in short-term returns or are influenced by competing crises (wars, recessions, pandemics).<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If a project loses momentum halfway through, it can stall for years before being revived under new leadership or funding. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Human Factors: Greed, Laziness, MistakesPeople might manipulate or \u201cgame\u201d a new system for personal gain, requiring additional layers of oversight, audits, or checks.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Large, error-free collaborations are extremely challenging. Miscommunications, internal power struggles, shifting alliances, and management flaws multiply delays. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
5. Learning \u201cAssembly and Fabrication Design Management\u201d for All<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
You mention focusing on something specific\u2014picking a target not out of personal bias but out of genuine potential to help many people\u2014and methodically applying a design-management approach. That\u2019s often the best path to:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
1) Clearly define the problem 2) Gather input from all major stakeholders 3) Use systematic design \/ simulation \/ modeling 4) Build small prototypes or pilot studies 5) Iterate and refine quickly 6) Scale and standardize<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
This approach can be done in a way that invites more transparent and inclusive decision-making, which helps mitigate suspicion or power-plays.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
6. Can We Overcome These Barriers Faster?<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
AI-Driven Coordination<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Modern AI tools can accelerate collaboration by providing shared transparent analyses and simulations.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
AI can highlight win\u2013win scenarios that might be overlooked by purely human negotiation.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Open-Source Collaboration<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
By opening up designs, standards, and research data to the world, you reduce the sense that one group \u201cowns\u201d the technology.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If the roadmap, cost, and potential benefits are openly visible, it\u2019s harder for one actor to quietly manipulate the project.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Participatory Governance<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Involving communities, local governments, NGOs, and industry from day one can diminish the \u201coutsider imposing solutions\u201d phenomenon and build trust.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Mechanisms that guarantee fair distribution of benefits (e.g., profit-sharing, local ownership, transparent oversight) help unify stakeholders.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Adaptive Regulation<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Governments can design \u201cdynamic\u201d regulatory frameworks that pivot as real-world evidence emerges, instead of waiting for perfect data or the next decade\u2019s rewrite of the law.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
This reduces the lag between the lab, pilot studies, and mainstream adoption.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
7. Summary<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Why Decades, Not Months?<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
It\u2019s not just about solving the technical constraints. Huge projects also must align with human psychology, politics, entrenched economic interests, legal frameworks, and intangible cultural factors. Aligning these can take far longer than the raw engineering or computational tasks.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Path Forward<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Pick a concrete and worthwhile goal (health, sustainability, new materials) that truly helps others.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Apply rigorous design-management methods\u2014iterative, transparent, inclusive.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Address stakeholder motivations systematically: fairness, shared benefits, open data, and adaptive regulations to reduce mistrust.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Leverage AI but remain aware that no amount of computational power fixes incentive misalignments by itself.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
The heart of the matter is that large, global, systemic transformations always face the \u201chuman dimension.\u201d Until that dimension is adequately structured and resolved, even the most advanced technology remains, as you say, \u201cnext year\u2019s solution\u201d\u2014always on the horizon, forever just out of reach.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
RichardKCollin2 said:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
What can AI groups do to help? What can the emerging intelligent entities do to help?<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Most importantly how can they structure relations with all human and AI society so they are trusted, trustworthy, accessible, reliable and help everyone? <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
At one time “banks” and “advisors” were neutral and trustworthy. Wrapping them in circuits, algorithms, software and machine has not made that role in society easy to evaluate and judge.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Unless the roles and responsibilities and benefits are clear, it won’t have a chance to even begin.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Here is OpenAI ChatGTP Plus running ChatGPT o1 says:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Thought about AI’s role in society for 7 seconds<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Below is an outline of practical steps AI groups (and emerging \u201cintelligent entities\u201d at large) can take to become trusted, trustworthy, and reliable partners to human society. For each aspect, we\u2019ll consider structural, technical, and cultural approaches\u2014because trust ultimately depends on transparent governance, clear responsibilities, and demonstrable benefits for all stakeholders.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
1. Clarify Roles, Responsibilities, and Benefits<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Role Definition<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
AI groups should explicitly define:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Their purpose (e.g., \u201cWe develop decision-support tools for healthcare.\u201d)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Their scope (what they do and, importantly, what they do not do)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Their obligations to the public, users, employees, and regulators<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Benefits and responsibilities must be clearly stated in publicly available, understandable terms.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Alignment with Societal Goals<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Demonstrate how the group\u2019s work directly connects to problems people care about:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Healthcare access<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Environmental sustainability<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Education<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Financial inclusion<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Make it explicit: \u201cHere is the problem we\u2019re solving, here is how we\u2019ll measure improvement, and here is who benefits.\u201d<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Multi-Stakeholder Involvement<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
At an early stage, invite representatives from diverse groups\u2014consumer advocates, domain experts, local communities, regulatory bodies\u2014to co-create the AI\u2019s objectives or use cases.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If people see how the system is designed from day one, trust starts from familiarity, not from marketing hype.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
2. Adopt Transparent, Accountable Governance<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Open Governance Model<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Publish an organizational governance framework describing how decisions are made:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Decision_Pipeline = f(technical_teams, ethics_panel, user_feedback, <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
regulatory_requirements, board_oversight)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Have an independent ethics board or advisory council with actual veto or revision power on critical design decisions.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Clear Mechanisms of Accountability<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Internal Audits: Set up formal processes that regularly review code, data, decisions, and outcomes.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
External Audits: Allow third-party organizations or \u201cred teams\u201d to evaluate the AI\u2019s performance and risk factors.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Liability Frameworks: Publicly define how your group handles mistakes or misuse. If you produce a system that harms people, who is responsible, and how are they made whole?<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Explainable Outcomes<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
For any AI that impacts crucial human decisions\u2014loans, medical diagnoses, hiring\u2014provide explainable components:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Summaries of the factors leading to a decision<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Mechanisms for appeals or second opinions<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Trust grows when users (and regulators) can see why a model does what it does.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
3. Foster Broad Accessibility and Inclusion<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Equitable Access to AI Tools<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Make sure the AI outputs or services are not locked behind paywalls or exclusive partnerships\u2014especially if it\u2019s addressing public or essential goods (health, education).<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Offer tiered or open licensing so nonprofits, small companies, and developing regions can also benefit.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Localized Interfaces and Language Support<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
AI systems should be linguistically and culturally adaptable.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If the AI is a medical assistant, for example, ensure it can handle local contexts, languages, and cultural norms.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
User-Friendly Communication<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Provide user interfaces and documentation that non-technical people can grasp.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
This includes clarifying limitations: \u201cThe system\u2019s best performance is on tasks X and Y, but it may be less accurate on tasks Z.\u201d<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
4. Ensure Robust Reliability and Security<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Resilience by Design<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Build systems to handle unexpected inputs and gracefully degrade rather than producing catastrophic errors.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
For example:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
AI_Safety_Functions = f(redundant_models, anomaly_detection, sandbox_testing)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Perform frequent \u201cstress tests,\u201d akin to banking stress tests, to see how the AI behaves under extreme or malicious scenarios.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Data Privacy and Protection<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Encrypt sensitive data.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Comply with global privacy regulations (GDPR, etc.) and seek to exceed them, demonstrating genuine care for user protection.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Provide transparent data governance so everyone knows what data is collected, how it\u2019s used, who can access it, and how long it\u2019s retained.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Continuous Monitoring and Updates<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
The environment changes, attacks evolve, and training data can become outdated.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Establish lifecycle maintenance to:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Retrain or calibrate models periodically<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Patch vulnerabilities proactively<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Announce changes transparently<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
5. Cultivate Neutrality and Fairness<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Bias Audits<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Regularly run statistical bias checks and measure:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Disparate_Impact ~ f(user_demographics, model_predictions, real_outcomes)<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If the AI systematically under-serves or misjudges certain groups, incorporate corrective measures.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Open Datasets for Validation<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Provide test sets or partial (anonymized) real-world data for external checks and research.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Encourage data scientists or academics to replicate findings, revealing any hidden biases or performance blind spots.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Transparent Model Objectives<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Communicate your system\u2019s utility function: what is it optimizing (accuracy, fairness, revenue), and how do these trade-offs work?<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If there are conflicts (e.g., maximizing profit vs. ensuring equitable outcomes), publicly acknowledge them and describe how you resolve them.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
6. Engage in Collaborative, Ethical AI Ecosystems<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Open-Source or Shared Knowledge<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Contribute to open-source libraries, publish technical insights, or share best practices.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
This builds a broader trust ecosystem, where improvements to safety, interpretability, or fairness benefit everyone.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Standards and Consortiums<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Join or form multi-stakeholder consortiums to set industry-wide standards:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Model interpretability best practices<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Safety benchmarks<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Data-sharing norms<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
When AI groups shape and follow common rules, it reduces suspicion of \u201csecret agendas.\u201d<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Public Education and Outreach<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Host workshops or educational programs that demystify AI for students, professionals, and the public.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Transparent engagement can turn confusion or fear into understanding and constructive critique.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
7. Long-Term Stewardship and Evolution<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Adapt to Societal Needs<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Keep an ear to the ground for shifting social priorities (e.g., climate crises, pandemics, new economic realities) and adapt your AI offerings accordingly.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Show that you\u2019re responsive rather than locked to outdated business models.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Ethical and Human-Centered AI Research<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Invest in fundamental R&D about AI alignment, meaning an AI\u2019s objectives remain compatible with widely accepted human values or the explicit instructions of its user community.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If advanced AIs act more autonomously, develop fail-safes and rigorous alignment protocols.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Proactive Policy and Self-Regulation<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Instead of waiting for external forces to impose rules, AI groups can self-regulate by adopting strong internal policies or pledging to adhere to known frameworks (e.g., the OECD AI Principles).<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Leading the compliance conversation fosters trust and sets an example for newcomers.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Synthesis: A Trustworthy AI Society<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
For AI to gain the status once held by \u201cneutral\u201d banks or advisors, it must demonstrably serve the public good and prove it can safeguard interests fairly and consistently. The essential elements are:<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Transparency about data, processes, and motivations<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Accountability with clearly defined responsibilities, audits, and liability<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Inclusivity so that all stakeholders (not just profit-driven entities) have influence and benefit<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Reliability in delivering services with consistent quality and safety<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Continuous Evolution to adapt to new insights, social demands, and technological risks<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
If AI systems and their creators follow these principles, they stand a far better chance of being perceived as trusted, neutral, and beneficial. This will not happen automatically\u2014intense diligence, community engagement, and open collaboration are required. But these steps give a solid starting structure for AI groups (and emerging intelligent agents) to earn genuine trust, serve everyone\u2019s interests, and demonstrate their worth to human society.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
<\/div>\n","protected":false},"excerpt":{"rendered":"

Did you try this? Fast accurate calculations are helpful, but the space of possible things to investigate and try is still large. For a variety of reasons, “small parts can be stronger”, “simple shapes take less time”, “many bonds can make a mess” – it makes sense to assemble and use molecular parts, rather than
Read More »<\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[73],"tags":[],"class_list":["post-19190","post","type-post","status-publish","format-standard","hentry","category-all-knowledge"],"_links":{"self":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts\/19190","targetHints":{"allow":["GET"]}}],"collection":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=19190"}],"version-history":[{"count":2,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/19190\/revisions"}],"predecessor-version":[{"id":19192,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/19190\/revisions\/19192"}],"wp:attachment":[{"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=19190"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=19190"},{"taxonomy":"post_tag","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=19190"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}