The Frankenstein Dilemma: Can We Control the AI Monster We've Created?

When Elon Musk and Steve Wozniak team up to sound the alarm about the dangers of a cutthroat race to conquer the AI world, it's time to sit up and take notice. Last week, I posed a question on this topic in my newsletter (follow me on LinkedIn at https://www.linkedin.com/build-relation/newsletter-follow?entityUrn=7044762993154289664), and now, it's a hot topic. But as always, I'll cut to the chase and focus on the issue at hand rather than get sidetracked by the motives of those who signed the letter. While some may have a long history of championing ethical AI, others could have a hidden agenda, so let's stick to the facts.

Nothing new under the sun.

Whether we can trust AI is hardly new, but it remains essential as we grapple with the ethical considerations of technological advancement. For all our modern inventions, the fundamental mechanisms that drive business models (and related ethical questions) have remained remarkably consistent throughout history. For example, eBay's Dutch auction system is not dissimilar to the tulip trade that flourished in 17th-century Holland, while Google's advertising algorithms serve the same purpose as 19th-century French advertising agencies[1]. But while the tools may have changed, the underlying ethical questions remain the same. We must consider them seriously as we navigate new and powerful technologies.

Humans have always been wary of progress and its impact on our moral compass. From Sophocles' Antigone[2], in which the invention of new tools is seen as a double-edged sword[3], to Rousseau's rejection of the idea that scientific progress leads to moral progress during the Enlightenment, to the skepticism expressed by Diderot, D'Alembert (even if they edited the Encyclopédie), and Voltaire (speaking of decadence), we have long grappled with the tension between innovation and ethics. Modern writers like Asimov, Orwell[4], and Bradbury have similarly pondered the impact of technological change on our humanity. As Asimov himself said[5], "Science fiction can be defined as that branch of literature which deals with the reaction of human beings to changes in science and technology."

In the end, trust remains central to our relationship with hashtag#ai. Therefore, we must stay vigilant and thoughtful as we navigate this new terrain, using our ethical compass to build a relationship with this powerful, transformative technology.

We need to build trust with AI products... and their designers.

"Ought I to trust or not?" is the epistemological question at the core of this dilemma. But what is trust, really? It's a leap of faith, a risk mitigated by our ability to determine if someone or something is trustworthy. Therefore, to trust AI, we must be able to assess its trustworthiness.

Of course, a significant part of the trustworthiness depends on the AI-driven products themselves. Machine learning models must be explainable. We need to know the data used during the training phase and the biases that the data may contain. Algorithms may be opaque to the average person, but transparency is essential for trust. We need processes in place, just like in other complex scientific domains, such as medicine and food, to ensure the system is trustworthy. Ultimately, we can use hashtag#ai to unbias hashtag#ai, or another technological innovation, like hashtag#blockchain, to create trust in hashtag#ai.

However, the other critical part of trust comes from us, and that's where education comes in. We must learn to ask questions about data, understand the limitations of algorithms, and correctly use technology. We should have computer science and mathematics knowledge but also learn about human beings and our history. We should always be curious and never tire of asking why. Only then can we agree with Hemingway's famous quote: "The best way to find out if you can trust somebody is to trust them."

Education also has the advantage of teaching us how to use hashtag#ai properly. This is no easy feat, as evidenced by the debates about the accuracy of language models like hashtag#chatgpt. But we need to remember that a language model is just an interface between forms of communication, not a search algorithm that needs to be relevant or true. We must not fall into the trap of believing that what hashtag#ai generates is always appropriate, just like we cannot assume that the top five sites that Google gives us in response to a query are valid and accurate in content. Building trust in hashtag#ai is much like building a team of Navy Seals, where trust is built over time and through rigorous testing.

When designing hashtag#ai inventions, we must consider their usage at every stage, from advertising to using to selling or trashing. They must be trustworthy throughout all interactions with users. This creates a triangle of trust between the user, the product, and the company building it, where the image and legitimacy of a company reflect on the product.

In the end, trust in hashtag#ai depends on both the product and the user, and only with a culture of trust and education can we build a better future for hashtag#ai.

AI products inherit the personality traits of their companies.

In the world of hashtag#ai , it's not just about technology anymore. It's about the personality of the company behind it. According to Jennifer Aaker's brand personality theory, how users perceive a company's brand affects how they trust its AI product. Take the case of Mercedes-Benz AG, for example. When the car company faced a crisis in 1997 with the class A moose test scandal, it had to recall 17,000 cars and spend millions to fix the issue. But because Mercedes-Benz AG' brand personality was built around precision, discipline, and efficiency, it couldn't win the media battle. A brand with a more rugged and adventurous personality may have fared better.

Fast forward to today's AI landscape, and we see a race between tech giants like OpenAI, Microsoft, and Google to create the best language models. OpenAI was a relative unknown just a few months ago, but its hashtag#chatgpt model has put it in the spotlight. Its brand personality is all about excitement and innovation. Google, on the other hand, has a lot more at stake. Its brand personality is built around competence, reliability, and trustworthiness.

In the high-stakes game of hashtag#ai, even competent tech giants like Google can make costly blunders, though. The search engine behemoth launched its hashtag#bard language model early, promptly delivering a factual error in a widely shared demo. The result? A $100 billion loss in Alphabet Inc. shares.

But the competition hasn't slowed down. Microsoft has poured billions into its partnership with OpenAI, securing its place as the exclusive cloud provider for its unique hashtag#chatgpt algorithm. But with that partnership comes a question: how trustworthy is hashtag#chatgpt now, and how will Microsoft's brand personality impact its image in the future? It's a question of trust; like market share, trust is a fickle mistress. It takes time to build and can crumble instantly, leaving even the most prominent players in the industry vulnerable to the whims of public opinion.

Enlightening the users is a way to gain trust.

Only a few products will be game changers with the rise of artificial intelligence products. Which ones? The answer is simple: trust the ones that help us become better at what we do, not the ones that replace us.

Should we trust the products giving us an optimal result? Optimization, the mathematical pursuit of truth, may seem like the ideal solution to every problem. However, nothing is more doubtful than pure efficiency. Who can say that one solution is the best no matter what? It is too complicated to understand how machine learning built such a solution fulfilling tens of constraints. What seems undebatable becomes doubtful and then dangerous. Indeed, the truth is that people make mistakes, and optimization alone can't account for the many variables and uncertainties we face daily.

That's where a trustworthy hashtag#ai comes in. By providing alternatives (not optimality) and assisting us in making more objective and informed decisions, hashtag#ai can help us use our expertise to its fullest potential. And the best part? We can still make mistakes, but they'll be less frequent, less costly, and less dangerous.

At Quincus, my team and I are building a hashtag#metaverse of the hashtag#logistics chain, using hashtag#ai to simulate worldwide logistical operations. But we're not just focused on providing the optimal solution to every problem. For example, we know that drivers and dispatchers are often faced with incomplete information and tough decisions. That's why we provide them with more data, faster, and a range of alternatives to weigh and consider.

By exploring and trying out different solutions, drivers and dispatchers can become better at what they do instead of consistently relying on the same old solutions. But, of course, too many alternatives can be overwhelming, just like a dense menu at a restaurant. So, we strike a balance in our products to provide just enough alternatives to be helpful without causing decision paralysis.

Last question: How should we build hashtag#ai products we view as tools to augment human capabilities, not replace them?

The design of everyday AI things.

Designing AI products that people can trust is no easy feat. But by focusing on the user experience and leveraging existing products' goodwill, we can build trust with users.

First and foremost, we must consider the experience of using machine learning. As an AI product designer, I prioritize making user-centric prototypes, ensuring users have a pleasurable experience they can trust. Never find refuge in functionalities (any scientist's comfort zone) and focus on experiences.

Secondly, I use Don Norman's book to adapt notions of #affordances and #signifiers to AI research and innovation. Affordances represent the possibilities for how a person can interact with AI products. Even if not all are visible explicitly or intuitively, all are crucial for building trust. Signifiers, on the other hand, are critical in AI for explainability. With a signal's simplicity, they counterbalance the complexity of what happens inside neurons' layers. In other words, I transform the need for the explainability of AI into the concepts of coverability and understanding. Users must discover all the possible actions of the product alone and understand what the outputs the machine sends mean. If you have played once with Google's TensorFlow playground, you will immediately understand that even complex mathematical principles that govern deep learning can be intuitive.

When designing hashtag#ai products, it's also crucial to leverage the goodwill of existing products (hashtag#ml-driven or not) by bringing back their affordances and signifiers into the hashtag#ai design.

Finally, feedback is crucial for building trust. hashtag#ai products need to talk to their users, explain what happens, and learn from launch after launch, use after use.

Designing trustworthy hashtag#ai products requires prioritizing the user experience, leveraging existing products' goodwill, and incorporating feedback mechanisms. With these strategies in place, we can build trust with users and create hashtag#ai products that people can rely on.

Can we trust AI? Let's learn to design trustworthy ML products.

In conclusion, building trust in hashtag#ai is crucial for its successful adoption and growth. While the debate on the ethics and risks of hashtag#ai is not new, the rapid advancement of technology has brought this issue to the forefront. To build trust, we need transparency, accountability, and education. We must ensure that hashtag#ai-driven products are explainable, unbiased, and trustworthy while educating users about the limitations and appropriate use of technology. We must also consider the personality of the companies building hashtag#ai products, as their brand perception affects how users perceive their products. Ultimately, with a culture of trust and education, we can create a better future for hashtag#ai.

Notes and References

[1] e.g., the Dollingen Advertising Agency (1841) and then the General Advertising Company (1845)

[2] Sophocles, Antigone: https://ir.canterbury.ac.nz/handle/10092/9681

[3] "Possessing a means of invention, a skillfulness beyond expectation, now toward evil he moves, now toward good."

[4] Orwell G., The Collected Essays, Journalism, and Letters of George Orwell, Volume IV, In Front of Your Nose 1945-1950

"Man needs warmth, society, leisure, comfort and security: he also needs solitude, creative work and the sense of wonder. If he recognised this he could use the products of science and industrialism eclectically, applying always the same test: does this make me more human or less human?"

[5] Asimov I., "How Easy to See the Future!", Natural History, 1975

[6] Aaker, J.L. Brand personality: conceptualization, measurement and underlying psychological mechanisms. Stanford University, 1995.