What it Means to Be Human: Our Competitive Advantages Over AI

I recently wrote this article for my blog, in which I explore the skills that experts believe will be the biggest hurdles for artificial intelligence in the coming decades. Do you agree? I would love your feedback.
{link removed by moderator}

1 Like

is it my imagination, or is your blog monetized with embedded advertising? And this link above is driving revenue for you? :thinking:

Can you not tell us a summary of your thoughts here?

Call me cynical, but after 5 years and thousands of hours of time donated to this community, when I see a person’s first post is an external link to a URL they make money from, it makes me suspicious of motive. I know it’s a character flaw, I’m working on it.

5 Likes

When you are non trusting it means you have trust in your doubt. There is no lack of trust just the trust is in the doubt, it means trust in your intelligence. And dont change your nature, it is natural after all​:grin:

Apologies. I’m new here so not used to the common behaviors. Sorry for posting. There are ads on the site, which is an add-on widget I included…that has earned me all of $0.02. One does not invest so much time and energy into research for pennies, so it is an honest attempt to gain feedback. If the common way of communication is summaries, I will attempt to do so.

As requested a summary of my article:

My article discusses the potential advantages humans have over AI in the 21st century, drawing from the ideas of experts like Kaif-Fu Li, Yuval Noah Harari, and others. Li’s matrix identifies jobs at risk due to AI advancements by looking at their social component and focus on optimization or creativity. Roles in the “danger zone” are those with low human interaction and repetitive processes, while those in the “safe zone” require successful human relationships and creative or strategic thinking. However, Li and Harari believe that AI still has a long way to go before reaching the “singularity,” where it surpasses human intelligence in significant ways.

Humans have a competitive edge over AI in areas such as emotional intelligence, critical thinking, and creativity. These skills allow humans to solve problems relevant to human existence more effectively than AI can. Emotional intelligence makes humans capable of understanding and being motivated by feelings, while critical thinking helps evaluate situations and make decisions based on inferences. Creativity, driven by individual experiences and cultural backgrounds, allows humans to generate unique ideas that contribute to a vast pool of collective creativity.

I conclude by acknowledging that a future with advanced AI and seamless human-non-human interaction is possible, but it is not imminent. For now, humans maintain an advantage in complex areas such as leadership, strategic thinking, problem-solving, and ethics, which enables them to move forward and address issues more effectively than AI. The focus should be on nurturing these qualities and ensuring that humans continue to excel in these areas as AI technology progresses.

The “tears in rain” scene from Blade Runner mentioned at the end is a vision of the future, perhaps a far-off future.

2 Likes

Hi @Roy_Tomizawa dont take it personally its just non-artificial inteligence in action​:grin: i.e. doubting, evaluating…

To me artificial intelligence is like an artificial flower, basically not a flower. I dont want to go into much details about it but think about concioussness and if ever a computer can have concioussness and how a human or any organism becomes conciouss.

I would dare to disagree :slight_smile:

At least the way I see it, humans overvalue the traits that they are “proud” of (like playing chess was considered a great achievement of human mind (chimps couldn’t do it)). But the areas the AI excels is exactly there I think.

On the other hand, I think professions like construction workers are the last ones to be replaced. (For current humanoids even making a cop of coffee is a big deal). Self-driving received billions of investment and I would dare to state that the progress was not adequate. So learning to navigate the complex environment and not causing havoc is a big hill to climb (pun intended).

Just my thoughts :slight_smile:

1 Like

He can use ChatGPT for summarizing… :smile:

2 Likes

He should have done it and specified below 50 words​:relieved::grin:

2 Likes

Tough crowd… ; ) Seriously, I do appreciate the honest and open feedback! Thank you all.

3 Likes

I tried a quick and dirty version using code from the short course…here ya go

prompt = f"""
Your task is to generate a short summary of a this article \
 from a web site. 

Summarize the review below, delimited by triple 
backticks, in at most 50 words.

The article discusses the unique qualities that make humans superior to AI, including creativity, empathy, and intuition. It argues that these qualities will continue to be valuable in the workforce, even as AI technology advances.

Here’s the 100 word version…

The competitive advantages humans have over AI are empathy, creativity, and social skills, according to a new report from the McKinsey Global Institute. The report found that while AI can outperform humans in certain tasks, it lacks the ability to understand human emotions and social cues, and is not capable of generating truly original ideas. The report suggests that humans should focus on developing these uniquely human skills in order to stay relevant in the workforce as AI continues to advance.

EDIT: NOTE: this is only from the first 4000 characters, as my unpaid OpenAI account has a character limit on what it will agree to summarize

FURTHER EDIT:

I tinkered around with HTML tag stripping. Turns out the vast majority of that page is links and style sheet formatting. With all of that stripped out, a free account API KEY and the code from the Summarizing lecture of the prompt engineering Short Course was able to generate this summary (expanded to 200 words):

Humans still have a significant edge over artificial intelligence (AI) in leadership, strategic thinking, problem-solving and ethics, according to a review by Forbes. While AI can be programmed to be curious, to ask questions and evaluate, humans are still more effective in providing relevant explanations and solutions in the face of complex human conundrums. The review also highlighted that humans are consistently superior at the skills of emotional intelligence, critical thinking and creativity, which powerfully enable our ability to move forward and solve issues, particularly those relevant to human existence. The review cited Kaif-Fu Li, a Taiwanese computer scientist and venture capitalist, who explained that there are two axes by which we can identify what kind of jobs are at risk due to advances in AI: whether a role has a high or low social component, and whether a role focuses on optimisation at one end, or creativity or strategy at the other.

From what I can tell from a quick read on my own, the article heavily leverages and repackages work originally published by Forbes and Kaif-Fu Li.

@JavierSC we had the same idea :slight_smile:

1 Like

I’m deleting the link from the OP, because I believe it violates the Code of Conduct.

I notice this is one of the takeaways from the cited McKinsey report. I think this has already been disproven in the real world. Just in the last couple of days I read about an AI trained to compete in some boat racing game. It quickly learned on its own that it could get the highest score by attacking other racers, causing them to be swamped by waves, or cause them to catch fire, none of which was it explicitly told about or trained to do. Seems pretty human to me.

Thanks you sir for the summary :grin:, the 100 words is better. Now I am clear whats this all about​:slightly_smiling_face:

1 Like

At this point I should say it is probably some combination of pieces of what it was fed and learned from, but still I dont think we can call it this combination an authentic new forthbringing. Otherwise it would defeat the very learning process itself, where did the model get it, it learns from some data that it is fed after all…

@ai_curious
I tried a quick and dirty version using code from the short course…here ya go

prompt = f"""
Your task is to generate a short summary of a this article \
 from a web site. 

Summarize the review below, delimited by triple 
backticks, in at most 50 words.

This is really an example of how to apply the new knowledge hahaha

Does seem untrue, but only for Reinforcement Learning approaches, IMO. I haven’t yet seen like an NLP application that introduces truly new ideas (unless you count confabulation aka hallucinations lol)…

Something that is perceived as a ‘perfect’ flower. What is it?