A Shift in Paradigm

This technology is here to remain.

Despite indications that AI is entering a financial bubble, it’s hard to imagine a future where Generative AI products like ChatGPT, DALL-E, or Claude are not an integral part of our daily lives, whether we embrace them or not.

However, a recent announcement heralds a new phase for AI—one that transcends mere productivity and business applications, invoking a modern interpretation of Machiavelli's principle that “the end justifies the means.”

Claude, the Constitutionally Guided Chatbot

Curious about how much funding Anthropic AI secured in its inaugural year?

A staggering $580 million.

This was achieved without an established product-market fit, revenues, or even a functioning product until quite recently.

Yet, this did not deter Google from investing $300 million at a valuation of $3 billion.

The Motivation?

The concept of Constitutional AI, an AI designed for a better future.

Should Humans Have the Final Say?

Anthropic AI's research introduces an innovative method for training AI that blends supervised and reinforcement learning in a unique manner.

At first glance, the approach may seem overwhelming, but it can be distilled into straightforward steps:

• Initially, they gather responses to harmful prompts from a baseline model.

> Unsurprisingly, these results are often harmful and toxic, given that the training data tends to be biased.

• The next step involves a ‘Response, Critique, Revision’ process, where the model reviews its own replies against a 'Constitution' created by humans that outlines principles and guidelines it should adhere to—a moral compass of sorts.

> This Constitution is the sole point of human intervention aside from the sampling phase.

• The result is a supervised model that is then utilized in a subsequent process where it evaluates generated responses against the Constitution without human input on what constitutes harm.

> In essence, at this stage, the model is not trained by human feedback but is guided by its own interpretation of harmfulness.

> Both fascinating and alarming!

• Ultimately, by comparing responses, they generate a preference model that rates harmfulness and combines it with the initial model to create a finely tuned version.

The outcome? A groundbreaking AI model that, by using its own understanding of harm, aims to minimize negative impacts while providing responses in the most harmless manner.

This seems like a noble pursuit, doesn’t it?

However, the reality is more complex—some may prefer to collaborate with less ethical entities than to partner with Anthropic AI.

What’s the Reason Behind This Hesitance?

It relates to the Effective Altruism movement.

Effective Altruism: A Double-Edged Sword?

Google's investment now extends beyond AI alone.

Anthropic AI is closely associated with the Effective Altruism (EA) movement, which focuses on maximizing goodness for the world.

Their mission is not just to improve conditions for the living, but to make sacrifices for the benefit of future generations.

Notably, among EA advocates is Sam Bankman-Fried, the former CEO of the now-defunct FTX Crypto exchange, which happens to be a significant investor in Anthropic AI.

Coincidence?

This movement, while appealing, has sparked fear due to its extreme views on what it means to do ‘good.’

Immanuel Kant would likely endorse it.

A major focal point for EA is AI safety. As Will MacAskill, the movement's founder, noted, one danger of AI lies in its potential to create intelligent agents that surpass human capabilities.

Yet, it’s the underlying philosophy that raises concerns, illustrated by a simple dilemma:

Which Path to Take: Banker or Volunteer?

If your aim is to maximize global good, should you pursue a career as an investment banker or volunteer for an NGO?

Initially, one might assume that volunteering is the better option. However, according to the Effective Altruism framework, that would be a misguided choice.

The ideal route, they argue, is to become an investment banker, accumulate wealth, and then donate to altruistic causes.

This rationale, while logical, raises ethical questions.

This perspective mirrors Sam Bankman-Fried’s vision of amassing wealth only to redistribute it.

From a purely economic standpoint, this argument appears sound... or does it?

The End Justifies the Means

Napoleon Bonaparte was known for his admiration of ancient texts, particularly Machiavelli’s works.

He famously inscribed “the end justifies the means” in his copy of The Prince, a phrase many mistakenly attribute to him rather than its original author.

This debate has persisted for ages, and I can't help but notice how closely aligned the Effective Altruism approach is to this idea.

While I appreciate their intentions, it can lead to justifying injustices against a minority for the alleged greater good of the majority.

Yet, even if you align with Napoleon's thinking, the rationale behind Google's investment in Anthropic is not immediately clear.

Google: A Leader in AI

Google Research stands as the world's foremost AI research entity.

This doesn’t necessarily guarantee that its AI products will excel, as research and commercial applications are distinct.

However, it's likely that Google’s proprietary language model, LaMDa, will also perform exceptionally well.

Moreover, Google's parent company, Alphabet, owns DeepMind, which is developing Sparrow, another promising chatbot set to enter the market.

This leads us to question: why invest in Anthropic AI?

What unique advantage does this nascent company possess that compelled Google to invest a staggering $300 million in Claude?

Constitutional AI: A Game Changer?

Given that Anthropic AI champions the idea of Constitutional AI, Google may believe that this training methodology represents the most effective approach for large language models.

Perhaps, although still unproven, distancing human biases from defining what is harmful could indeed be the optimal way to train these models, making Claude the most advanced chatbot available.

Or, there may be even deeper motivations behind Google's involvement, pursuits that transcend AI and seek to answer a more profound question:

Are we truly crafting an AI capable of making ethically sound decisions that surpass human reasoning without interference, or are we inadvertently creating a threat?

A Final Thought

If you've engaged with this article, you're already ahead of 95% of the population regarding AI understanding. However, that still leaves many at your level.

What if you aim to surpass 99% of society? That would elevate your understanding.

If that's your ambition, I have exciting news.

By subscribing to my weekly newsletter, you'll explore intricate topics on innovation in an easily digestible format, keeping you updated on the latest in Crypto and AI.

I can’t guarantee riches, but I can assure you that knowledge is invaluable, and true wealth is rooted in understanding.

> Stay informed about the technologies shaping our future. Prepare to capitalize on that knowledge.

TheTechOasis: Simplifying AI & Crypto

For more insights, check out the landing page for The Tech Oasis newsletter, designed to make AI and Crypto accessible to everyone, regardless of technical expertise.

www.thetechoasis.com

You also have the opportunity to become a Medium member, gaining access to a wealth of articles on any topic that piques your interest while supporting this writer.

Join now! Don’t hesitate to elevate your understanding.