Opinion

Lets retire this hackneyed term: while ChatGPT is good at pattern-matching, the human mind does so much more

Thu 30 Mar 2023 10.55 EDT

Elon Musk and Apples co-founder Steve Wozniak have recently signed a letter calling for a six-month moratorium on the development of AI systems. The goal is to give society time to adapt to what the signatories describe as an AI summer, which they believe will ultimately benefit humanity, as long as the right guardrails are put in place. These guardrails include rigorously audited safety protocols.

It is a laudable goal, but there is an even better way to spend these six months: retiring the hackneyed label of artificial intelligence from public debate. The term belongs to the same scrapheap of history that includes iron curtain, domino theory and Sputnik moment. It survived the end of the cold war because of its allure for science fiction enthusiasts and investors. We can afford to hurt their feelings.

In reality, what we call artificial intelligence today is neither artificial nor intelligent. The early AI systems were heavily dominated by rules and programs, so some talk of artificiality was at least justified. But those of today, including everyones favourite, ChatGPT, draw their strength from the work of real humans: artists, musicians, programmers and writers whose creative and professional output is now appropriated in the name of saving civilisation. At best, this is non-artificial intelligence.

As for the intelligence part, the cold war imperatives that funded much of the early work in AI left a heavy imprint on how we understand it. We are talking about the kind of intelligence that would come in handy in a battle. For example, modern AIs strength lies in pattern-matching. Its hardly surprising given that one of the first military uses of neural networks the technology behind ChatGPT was to spot ships in aerial photographs.

However, many critics have pointed out that intelligence is not just about pattern-matching. Equally important is the ability to draw generalisations. Marcel Duchamps 1917 work of art Fountain is a prime example of this. Before Duchamps piece, a urinal was just a urinal. But, with a change of perspective, Duchamp turned it into a work of art. At that moment, he was generalising about art.

When we generalise, emotion overrides the entrenched and seemingly rational classifications of ideas and everyday objects. It suspends the usual, nearly machinic operations of pattern-matching. Not the kind of thing you want to do in the middle of a war.

Human intelligence is not one-dimensional. It rests on what the 20th-century Chilean psychoanalyst Ignacio Matte Blanco called bi-logic: a fusion of the static and timeless logic of formal reasoning and the contextual and highly dynamic logic of emotion. The former searches for differences; the latter is quick to erase them. Marcel Duchamps mind knew that the urinal belonged in a bathroom; his heart didnt. Bi-logic explains how we regroup mundane things in novel and insightful ways. We all do this not just Duchamp.

AI will never get there because machines cannot have a sense (rather than mere knowledge) of the past, the present and the future; of history, injury or nostalgia. Without that, theres no emotion, depriving bi-logic of one of its components. Thus, machines remain trapped in the singular formal logic. So there goes the intelligence part.

ChatGPT has its uses. It is a prediction engine that can also moonlight as an encyclopedia. When asked what the bottle rack, the snow shovel and the urinal have in common, it correctly answered that they are all everyday objects that Duchamp turned into art.

But when asked which of todays objects Duchamp would turn into art, it suggested: smartphones, electronic scooters and face masks. There is no hint of any genuine intelligence here. Its a well-run but predictable statistical machine.

The danger of continuing to use the term artificial intelligence is that it risks convincing us that the world runs on a singular logic: that of highly cognitive, cold-blooded rationalism. Many in Silicon Valley already believe that and they are busy rebuilding the world informed by that belief.

But the reason why tools like ChatGPT can do anything even remotely creative is because their training sets were produced by actually existing humans, with their complex emotions, anxieties and all. If we want such creativity to persist, we should also be funding the production of art, fiction and history not just data centres and machine learning.

Thats not at all where things point now. The ultimate risk of not retiring terms such as artificial intelligence is that they will render the creative work of intelligence invisible, while making the world more predictable and dumb.

So, instead of spending six months auditing the algorithms while we wait for the AI summer, we might as well go and reread Shakespeares A Midsummer Nights Dream. That will do so much more to increase the intelligence in our world.

Evgeny Morozov is the author of several books on technology and politics. His podcast The Santiago Boys, about the tech vision of former Chilean president Salvador Allende, is out this summer

Do you have an opinion on the issues raised in this article? If you would like to submit a response of up to 300 words by email to be considered for publication in our letters section, please click here.

{{topLeft}}

{{bottomLeft}}

{{topRight}}

{{bottomRight}}

{{.}}

Follow this link:
The problem with artificial intelligence? Its neither artificial nor intelligent - The Guardian

Its almost 20 years now since a socially awkward young computer science student set up a website for rating hot women.

Facemash, as Mark Zuckerberg called his creation, was shut down within days. But this crass teenage experiment was still, in retrospect, the first faltering step down a road to something even he couldnt possibly have foreseen at the time: a social media phenomenon now accused of unwittingly helping to polarise society, destabilise the democratic process, fuel hate speech and disseminate dangerous conspiracy theories around the globe, despite what providers insist have been their best attempts to stamp out the fire.

We couldnt have predicted then, and arguably still dont properly understand now, what impact Facebook or Twitter or Instagram or TikTok have had on teenage mental health. We couldnt have anticipated how life online would change our sense of self, blurring the line between private life and public content; didnt grasp until too late how algorithms developed to drive social media consumption would shape what we read or hear, and consequently how we think or feel. But if we couldnt have accurately predicted that from the start, with hindsight, there were surely moments along the road when the penny should have dropped.

Had governments not allowed the tech giants to race so far ahead of regulation, they might have saved themselves years of clearing up the resulting mess. But blinded by the riches the industry generated, and diverted by the pleasure its products have undoubtedly given along the way, we all missed the moment. The fear is that were about to do the same with something infinitely more powerful and unpredictable.

Artificial intelligence is arguably both the most exciting thing that has happened to humankind in generations key to magical, transformative breakthroughs in everything from medicine to productivity and the most frightening, given its potential to upend the existing social and economic order at breakneck speed.

This week some of the worlds leading AI experts called for a six-month pause on training the next wave of systems more powerful than the now famous ChatGPT-4 chatbot which has demonstrated an uncanny ability to communicate like a human in order to better understand the implications for humanity. They warn of an out-of-control race to develop and deploy ever more powerful digital minds that no one not even their creators can understand, predict or reliably control.

Shortly afterwards the British government published a white paper arguing that, on the contrary, Britain has only a brief window of around a year to get ahead in that race, and should adopt only the lightest of regulatory touches for fear of strangling the golden goose.

The UK wont have a new expert regulator governing what some think could become an extinction-level threat to humanity; instead, ministers will empower a bunch of overworked existing regulators to do what you might have hoped they were already doing, and scrutinise AIs impact on their sectors using a set of guiding principles that may be backed up at some unspecified point by legislation.

The whole thing smacks of a government desperate for economic growth at all costs and perhaps also for something resembling a Brexit bonus; if the EU treads its usual cautious regulatory path, Britain will position itself as the comparatively unfettered, gung-ho home of the AI pioneer.

The white paper mentions the jobs AI will undoubtedly create but skates over the ones it will eliminate and the social unrest that could follow. (Think of what the decline of coal, steel and manufacturing did to rust belt towns across Europe and the US, and how that fuelled the rise of populism; now imagine AI replacing a quarter of all work tasks worldwide, as predicted in a report by Rishi Sunaks old employer Goldman Sachs this week.)

Ministers stress the extraordinary breakthroughs possible in healthcare. But they have less to say about new forms of fraud or mass disinformation that could be perpetrated using AI tools capable of communicating convincingly like a human, or about how autonomous weaponry could be exploited by terrorists or rogue states. They dont talk nearly enough about what new rights humans might need to live alongside AI, including perhaps the legal right to know when an algorithm rather than a person was employed to sift our job application, refuse us a mortgage, fake what looks like an entirely authentic image or craft a flirty response on a dating app (yes, theres an AI application for that).

The risk of AI becoming sentient, or developing human feelings, remains relatively distant. But anyone who has ever got enraged by Twitter knows were already way past the point of algorithmic systems affecting humans feelings towards each other. Michelle Donelan, the new cabinet secretary responsible for tech, breezily assured the Sun this week that nonetheless AI wasnt something we should fear; the government had it all in hand. Feeling reassured? Me neither.

A global moratorium on AI development sadly seems unlikely, given we havent managed that kind of worldwide cooperation even against the existential threat from the climate crisis. But there has to be some way of avoiding what happened with social media: an initial free-for-all that made billions, followed eventually by an angry backlash and a doomed attempt to stuff genies back into bottles. Artificial intelligence develops, in part, by learning from its mistakes. Is it too much to ask that humans do the same?

View original post here:
This gung-ho government says we have nothing to fear from AI. Are you scared yet? | Gaby Hinsliff - The Guardian

UNESCOs Recommendation on the Ethics of Artificial Intelligence is the first global framework for the ethical use of artificial intelligence. It guides countries on how to maximize the benefits of AI and reduce the risks it entails. To this end, it contains values and principles, but also detailed policy recommendations in all relevant areas.

UNESCO is concerned by many of the ethical issues raised by these innovations, in particular discrimination and stereotyping, including the issue of gender inequality, but also the fight against disinformation, the right to privacy, the protection of personal data, and human and environmental rights.

Industry self-regulation is clearly not sufficient to avoid these ethical harms, which is why the Recommendation provides the tools to ensure that AI developments abide by the rule of law, avoiding harm, and ensuring that when harm is done, accountability and redressal mechanisms are at hand for those affected.

UNESCOs Recommendation places a Readiness Assessment tool at the core of its guidance to Member States. This tool enables countries to ascertain the competencies and skills required in the workforce to ensure robust regulation of the artificial intelligence sector. It also provides that the States report regularly on their progress and their practices in the field of artificial intelligence, in particular by submitting a periodic report every four years.

To this date, more than 40 countries in all regions of the world are already working with UNESCO to develop AI checks and balances at the national level, building on the Recommendation. UNESCO calls on all countries to join the movement it is leading to build an ethical AI. A progress report will be presented at the UNESCO Global Forum on the Ethics of Artificial Intelligence in Slovenia in December 2023.

Here is the original post:
Artificial Intelligence: UNESCO calls on all Governments to ... - UNESCO

With the rapid boom of artificial intelligence, or AI, in the last few months, universities and educational institutions have scrambled to figure out how to tackle academic dishonesty concerns, responsibly and ethically use and teach about AI, and learn more about the quickly evolving technology.

The reliability of the software, the potential pros and cons of artificial intelligence and the academic integrity concerns it raises have all quickly dominated academic discourse.

Director of the Texas A&M University Writing Center Nancy Vazquez says the quick evolution of ChatGPT has been an important topic of both formal and informal conversation among faculty and staff. Faculty are concerned AI may enable students to engage in dishonest academic behavior.

One of the concerns is that when you get a university degree, you assume that person is capable of doing certain things, they have certain skills and knowledge, but, potentially, AI clouds that, Vazquez said.

Even with tools such as Grammarly and Microsoft auto-predicting text, Vazquez said she always recommends consultants at the writing center make sure that students ensure their professors are okay with it.

Students thinking about using any type of AI for classes or an application need to check with the authorities, whether that is an instructor or a program theyre applying for What are the policies? Vazquez said.

Vazquez says there are a range of worries with ChatGPT including the legality of who owns the generated text that AI produces. Additionally, Vazquez said librarians have spent hours looking for sources that were generated by ChatGPT that dont actually exist. Early versions of ChatGpt offered premium versions after a limited free trial, which also raises questions of access as some students might have the use of tools that other students cant afford according to Vasquez.

I also think about AI being useful for generating routine writing that people do or even [planning] out your writing, or a writing schedule but also create a renaissance for the personality, human characteristics and creativity that we bring, Vazquez said.

Currently, the Writing Center makes reference to AI under the plagiarism statement in the course policies section of syllabi tailored toward professors who want to mention AI in their class.

According to the [A&M] definitions of academic misconduct, plagiarism is the appropriation of another person's ideas, processes, results or words without giving appropriate credit, Vazquez said. You should credit your use of anyone else's words, graphic images, or ideas using standard citation styles. AI text generators such as ChatGPT should not be used for any work for this class without explicit permission of the instructor and appropriate attribution.

Philosophy professor Kenny Easwaran, Ph.D, said AI mimics the neural network of the brain and is able to deal with language by associating and connecting information. However, it only has a limited memory that can look a few pages back up with GPT-4 compared to only paragraphs with GPT-2.

This is an associonist and connectionist neural network-based system, and this is one of the main criticisms that this is not going to get us real artificial intelligence, Easwaran said. It looks really good, but that is just because it can battle convincingly.

The associationist and connectionist thinking that artificial intelligence is able to do rather quickly is also something that humans do, Easwaran said. However, it is harder to replicate symbolic reasoning that is oftentimes slower and used for things like math according to Easwaran.

One of the things that we try to develop in higher education is to get people to use this slow and effortful symbolic reasoning, Easwaran said.

Easwaran said because ChatGPT pulls from information that people have given it, it writes based on recognized patterns and memorized information. However, it is not as good at understanding arguments or why premises either support or dont support a conclusion, Easwaran said.

It can do certain things like write a recommendation letter, or a memo but if you do something like original intellectual writing, it cant do that, Easwaran said.

Easwaran said he suspects that in the same way people use spreadsheets electronically now and nobody writes by hand anymore, in five to 10 years, AI is going to be a more normalized part of writing papers. Though he said he doesnt know how exactly it will evolve in the future, it is important for students to understand that in the same way that Wikipedia can be a good source of information, ChatGPT is not perfect and does have errors.

Link:
Is rapidly evolving artificial intelligence cause for concern? - Texas A&M The Battalion

Is AI and computer Vision two different domain or just two sides of the same coin?

Computer vision is a branch of artificial intelligence (AI) that enables computers and systems to extract useful information from digital photos, videos, and other visual inputs and to execute actions or make recommendations based on that information. If AI gives computers the ability to think, computer vision gives them the ability to see, observe, and comprehend.Human vision has an advantage over computer vision in that it has been around longer. With a lifetime of context, human sight has the advantage of learning how to distinguish between things, determine their distance from the viewer, determine whether they are moving, and determine whether an image is correct. Using cameras, data, and algorithms instead of retinas, optic nerves, and the visual cortex, computer vision teaches computers to execute similar tasks in much less time. A system trained to inspect items or monitor a production asset can swiftly outperform humans since it can examine thousands of products or processes per minute while spotting imperceptible flaws or problems.Energy, utilities, manufacturing, and the automobile industries all use computer vision, and the market is still expanding.

A lot of data is required for computer vision. It repeatedly executes analyses of the data until it can distinguish between things and recognize images. For instance, a computer needs to be fed a huge amount of tire photos and tire-related things to be trained to detect automotive tires. This is especially true of tires without any flaws. This is done using two key technologies: convolutional neural networks and deep learning, a sort of machine learning (CNN). With the use of algorithmic models, a computer can learn how to understand the context of visual input using machine learning. The computer will look at the data and educate itself to distinguish between different images if enough data is sent through the model. Instead of needing to be programmed to recognize an image, algorithms allow the machine to learn on its own.

By dissecting images into pixels with labels or tags, a CNN aids a machine learning or deep learning models ability to see. It creates predictions about what it is seeing by performing convolutions on the labels, which is a mathematical operation on two functions to create a third function. Until the predictions start to come true, the neural network conducts convolutions and evaluates the accuracy of its predictions repeatedly. Then, it is recognizing or views images similarly to how people do. Similar to how a human would perceive a picture from a distance, a CNN first recognizes sharp contours and basic forms before adding details as it iteratively tests its predictions. To comprehend individual images, a CNN is utilized. Like this, recurrent neural networks (RNNs) are employed in video applications to assist computers in comprehending the relationships between the images in a sequence of frames. Here are some applications of computer vision:

A dog, an apple, or a persons face are examples of images that can be classified using image classification. More specifically, it can correctly guess which class a given image belongs to. A social network corporation would want to utilize it, for instance, to automatically recognize and sort out offensive photographs shared by users.

To identify a specific class of image and then recognize and tabulate its existence in an image or video, object detection can employ image classification. Detecting damage on an assembly line or locating equipment that needs maintenance are a couple of examples.

After an object is found, it is followed or tracked. This operation is frequently carried out using real-time video streams or a series of sequentially taken pictures. For instance, autonomous vehicles must track moving things like pedestrians, other vehicles, and road infrastructure in addition to classifying and detecting them to avoid crashes and follow traffic regulations.

Instead of focusing on the metadata tags that are attached to the photos, content-based image retrieval employs computer vision to browse, search, and retrieve images from massive data repositories. Automatic picture annotation can be used in place of manual image tagging for this activity. These tasks can be used in digital asset management systems to improve search and retrieval precision.

Visit link:
Computer Vision Vs Artificial Intelligence. What Is the Difference? - Analytics Insight