OnPoint Subscriber Exclusive

The Big Picture brings together a range of PS commentaries to give readers a comprehensive understanding of topics in the news – and the deeper issues driving the news. The Big Question features concise contributor analysis and predictions on timely topics.

PS Commentators Respond: What Lies Beyond the AI Tipping Point?

Feb 9, 2023

Artificial intelligence is a contentious issue, to say the least. Some tout its potential to boost productivity, accelerate research and development, and improve health-care quality and delivery. Others warn that it will cause unemployment to skyrocket, exacerbate inequality, and generate vast ethical – even existential – risks.

In this Big Question, we ask Andrea Komlosy, Sami Mahroum, Stan Matwin, and Gabriela Ramos to assess the implications of the AI revolution that, in the wake of OpenAI’s release of ChatGPT, appears to be imminent, if not already here.

To continue reading, register now.

Subscribe now for unlimited access to everything PS has to offer.

As a registered user, you can enjoy more PS content every month – for free.

Already have an account? Log in

new comment has been posted. new comments have been posted.

0 Comments on this paragraph, 9 Comments on this article

James Pickles Feb 23, 2023

I tried ChatGPT on a few knotty scientific and philosophical questions. Its responses were superficial and were just a recasting of information that was already widely out on the web. There was no sign of what in human beings we call insight or creativity. If the information-processing world becomes dominated by ChatGPT and similar, then there will be just the same information going round in circles but no further insight or creativity. The implications? 1. It shows how superficial much human writing is. 2. It will give greater powers to those who do indeed show true insight or creativity, as they will be the ones who will generate the clearly new and therefore very sought-after ideas. HOWEVER it would not surprise me if AI one day is able to move to that level. As for the positive comments about autonomous vehicle by your commentator - where, except under constrained and limited environments, has this worked? It is not just legal and political limitations as mentioned - it's because it doesnt work in many real-life situations. And even if it works on some roads, it has to work on all roads where the vehicle is likely to go, in all weathers, and in all situations, before you can get rid of the human behind the wheel (OK it could say the weather is too bad to go out today, or the roads are too crowded, or the roads are too bad, but the threshold would need not to be too far off where a human driver would put it, to be acceptable).
A new reply to this comment has been posted.
Boris Azais Feb 13, 2023

Large language models and other AI image generators represent a major technological step forward for the creative sector, and a great opportunity for hobbyists. However, the hype (or drama) is overblown. Language models are not even close to true, general AI. They are extremely powerful tools designed and trained to simulate human text across any format.

Very impressive, and exciting possibilities, but it won't change the division of labor. For sure, more people will be able to produce content themselves and it might be at a higher quality than they could have done by themselves before. But above micro scale, companies will continue to use external vendors, graphists, copywriters, etc. to get things done at a professional level. What people buy is not just some image generated randomly, but a service that will get them a quality and tailored product for their specific need.

Creators will clearly use models to their fullest extent. But their clients will not double guess them simply because they discovered Stable Diffusion. The advent of powerful content generators will not change the division of labor: clients still have their own business to run.
A new reply to this comment has been posted.
Boris Azais Feb 13, 2023

1. Need to explain how companies could gain an "ever-greater control over our minds, bodies, and experiences" through conversational chatbots gathering personal data. What type of data exactly would enable such greater control? How does this apply to pharma/biotech who don't sell directly to patients, but to hospitals and wholesalers, and their products are prescribed by doctors.

2. What would a "patronizing, authoritarian, data-based existence" look like?
a) I can see the value for patients of making greater use of personal data - not doubt it would help in identifying early enough potential health problems, and thus intervene preventively, rather than too late as often the case today.
b) Not sure to understand what is an "authoritarian existence". Are you concerned about potential abuse of having access to an extremely large sets of personal data? Who exactly would abuse it? The government of some companies? If the latter, how exactly they turn authoritarian, and how does it change in any way the fact that they have no legal rights or actual control over the population.
c) Same question: what is a "patronizing existence"? Is it a concern over overeach by authorities empowered by such level of informaton, known risk factors, and thus temptation to treat citizens as being unable to take proper action based on available data?

How exactly all the above is in any way linked, triggered, or the result of the emergence of powerful AIs tools? Concerns over the possibility for a government to suspend fundamental rights during a health crisis didn't wait AIs to be raised. Throughout the pandemic, large protests were organized against vaccine mandates, and the obligation to wear a protective mask. Were these people protesting "the rise of a patronising, authoritarian, data-based existence"? But how come they didn't fall for the powerful chatbots and amount of personal data available to companies "set to gain ever-greater control over our minds, bodies and experiences. Should we conclude that antivax elements and other protesters didn't share our common "naiveté, lack of self-conto, and vulnerability to addiction?" Or more likely did they fall in the information bubble of antivax organizations, and other peddlers of conspiracy theories?

There are bigger forces and deeper passions in human societies than chatbots used to produce more targeted advertisements.
A new reply to this comment has been posted.
Curtis Carpenter Feb 13, 2023

I think that within a decade a massively trained AI system will be able to outperform my primary care physician in accuracy and reliability of diagnosis and prescribed treatment of what ails us to a undeniable degree. He already relies extensively on computer technology and medical databases, spending a good deal of his (expensive) time at his keyboard. So to know which way the wind blows regarding society's relationship to AI going forward, my advice would be: watch the AMA as a "canary in the mine shaft."
A new reply to this comment has been posted.
Boris Azais Feb 13, 2023

1. Do we know how many people use language assistants and chatbots, how often, and what's their opinion? Not sure about the seduction value of assistants or chatbots. These apps are still a long way from performing at acceptable level: They repeatedly fail to understand basic voice commands, or to provide the information you're looking for.

2. Not sure if any part of pharma and biotech companies' business model involves gathering and handling personal data. Maybe different in the US, but EU and national laws impose strict guidelines and obligations on how companies regarding any activity with personal data. Health data is even more stricly regulated. Patients' health data gathered in the course of a clinical trial follow a specific protection regime and the obligation to comply with Good Clinical Practices in performing R&D activities.(https://www.ema.europa.eu/en/ich-e6-r2-good-clinical-practice-scientific-guideline)
A new reply to this comment has been posted.
Boris Azais Feb 12, 2023

By their very design, large language models and/or their conversational interface cannot produce general AI. They may seem to "know" a lot (thanks to the massive amount of material they've trained on), but they don't "know what they know". For example, even if they can write a poem in the style of Yeats, they can only do it by mimicking one word after the other the "Yeats format" they picked up throught their training. They take the path "of greatest reward" on a step by step, and don't even have the ability to quality check their final output.
- Ashok Peer Feb 16, 2023
  
  Sir, I agree that AI can write a poem in Yeat’s style by mimicking from his many poems. It is also true that AI, unlike humans, doesn't know what it knows. But if an ordinary person is given a poem composed by AI in the masterly style of Yeats and Wordsworth without being told of the truth, will he be in a position to discriminate the poem from the one written by a poet ( human)? I feel it has become too difficult to distinguish between the two.
  - Boris Azais Feb 28, 2023
    
    Except for what's in front of us, about all the information we receive is curated one way or the other by some institution. The question is the amount of trust we put in the source and how important ("proximate") the information is for us. For example, "what's in the fridge?" is critically important if we want to eat tonight, thus we'll put an extra effort to validate it. On the other hand, "how far from Earth is the nearer pulsar?" is not that critical for most. The authenticity of a particular poem would seem to fall closer to the latter. As such, generative AI will add to the volume of irrelevant information. The key is to ensure population literacy on such tools, and not to trust what's not coming from trustworthy sources (which won't fully prevent misinformation, etc).
    A new reply to this comment has been posted.
  A new reply to this comment has been posted.
A new reply to this comment has been posted.
Boris Azais Feb 12, 2023

The concept of an "AI tipping point" (which seems a recurring thing across contributions) is of little value to discuss AI tools and their societal implications. There's no single "AI", but a variety of AI-based tools that perform very different tasks. Should different AI models have their own tipping point, or we're talking about general AI? When Deep Blue beat Kasparov at Chess, I guess it was a tipping point for Kasparov, if not really one for AI. Best to leave this "tipping point" to the media (drama and hype) and other futurists, etc.
A new reply to this comment has been posted.

https://prosyn.org/FaB26bZ;

OnPoint Subscriber Exclusive

PS Commentators Respond: What Lies Beyond the AI Tipping Point?

Featured in this Big Question

More
PS OnPoint

New Comment

Contact us

Log in/Register

OnPoint Subscriber Exclusive

PS Commentators Respond: What Lies Beyond the AI Tipping Point?

Featured in this Big Question

More PS OnPoint

Edit Newsletter Preferences

Set up Notification

More
PS OnPoint