{"id":70135,"date":"2023-02-24T06:00:00","date_gmt":"2023-02-24T05:00:00","guid":{"rendered":"https:\/\/www.futuribles.com\/chatgpt-promesses-et-risques\/"},"modified":"2024-08-28T16:53:27","modified_gmt":"2024-08-28T14:53:27","slug":"chatgpt-promesses-et-risques","status":"publish","type":"post","link":"https:\/\/www.futuribles.com\/en\/chatgpt-promesses-et-risques\/","title":{"rendered":"ChatGPT: Promises and Risks"},"content":{"rendered":"
\n
The ability of the internet industry giants to bring about \u201cdisruptive\u201d innovations shows no sign of waning. On 30 November 2022, OpenAI \u2014 a company co-founded by Elon Musk and Sam Altman, and supported by Microsoft \u2014 announced the release of ChatGPT-3, the application that everyone is currently talking about. It is an artificial intelligence (AI) chatbot with a generative ability (GPT stands for \u201cGenerative Predictive Transformer\u201d) that would pass the Turing test[1]<\/a><\/sup> with flying colours, and that amazes most of its users with the fluidity and quality of its responses. <\/h6>\n\n\n\n

ChatGPT describes itself<\/a> as being able to \u201cprovide factual answers, understand complex topics, solve problems, and perform tasks\u201d. These so-called LLMs (Large Language Models) absorb a vast corpus of texts from the web, notably from Wikipedia, <\/em>but also from many other websites, thanks to a neural architecture capable of processing 175 billion parameters. They record the words used and those that occur in close proximity, as well as the context in which they occur. On this basis, they generate word sequences according to statistical rules based on the context and the keywords of the question. The model is then optimised during a training phase under human supervision. Ultimately, this application has surprised even its creators with its performance.<\/p>\n\n\n\n

It is worth noting the considerable interest it has attracted. The opportunity offered to internet users to test it is said to have attracted more than one million volunteers in a few days, and there is now talk of 100 million test users. Meanwhile, the share price of OpenAI has soared, and the performance of this socio-technical system is making headlines. For example, ChatGPT earned an MBA from the prestigious Wharton School of the University of Pennsylvania, despite performing poorly in maths. On a more serious note, the US internet news site BuzzFeed[2]<\/a><\/sup> recently signed a multi-million dollar contract to provide content to Facebook and Instagram that is to be developed with the help of OpenAI products. This announcement gave the company a boost after a period of financial difficulties.<\/p>\n\n\n\n

Aside from the enthusiasm and the questions raised by this sociosystem, its \u201cdisruptive\u201d character deserves attention. I will touch on just a few dimensions of this disruptive quality: the internal dynamics of the technology industry and especially of the \u201cBig Five\u201d internet companies (Alphabet [Google], Amazon, Apple, Meta [Facebook], and Microsoft); the questions raised about the act of creation and its legal protection; and the impact of this technology on the sphere of knowledge, access to that knowledge, and its transmission.<\/p>\n\n\n\n

In the realm of technology, ChatGPT could, in the first instance, transform the economics of search engines. What is the point of using Google if ChatGPT can provide me with an immediately useful response? This is a direct threat to Google and its business model, which relies heavily on targeted advertising directly linked to user clicks. Google perceived the danger straight away, and is now announcing products that use AI, including Sparrow, powered by Deepmind, and Bard (launched on 7 February 2023), a conversational agent directly connected to the web, which is a direct competitor to ChatGPT. In a way, Amazon is also under threat, as the platform could be bypassed in favour of direct access to the desired product and its supplier. And even the Meta group (Facebook, Instagram) could be undermined by the ability of AI to present users with fictitious connections, indistinguishable from real internet users. By investing US$10 billion in OpenAI and securing the exclusive use of ChatGPT in its cloud offering, Microsoft is disrupting the existing balance between the internet industry giants, who have so far avoided head-on competition. The emergence of this technology is likely to bring about a new era of fierce competition between the giants of the internet sector.<\/p>\n\n\n\n

The world of creativity is also directly concerned by the emergence of ChatGPT. This is obvious in the case of literature and the media. The writing of newspaper articles and literary works alike will change, as will the very concept of editorial responsibility. All written productions may become raw material to be reworked by a new rewriting industry, and their authors may find themselves stripped of their rights. But the threat to creativity is broader than this, as similar generative models have been developed for creating images from text; DALL-E<\/a>, also developed by OpenAI, is the one that stands out against other applications such as Midjourney<\/a>. Both these systems raise the question of the plundering of existing works and of the threats they pose to intellectual property. As with the emergence of photography or, more recently, Wikipedia, the concept of creation needs to be partly redefined, and its protection reformulated.<\/p>\n\n\n\n

Beyond these practical issues, these developments are a matter of concern for the entire sphere of knowledge. The probabilistic approach to the linking of words does not provide any guarantees regarding the selection of sources, the AI\u2019s faithfulness to those sources, or even their traceability. ChatGPT does not cite its sources, and this is a problem: the knowledge it provides is unverifiable, and its developers encourage us to treat it with a critical distance. This appears to be a flaw in ChatGPT\u2019s relationship to knowledge, and raises the question of transparency with regard to sources, which is the basis of academic practice. Breaking away from this principle opens the way to all kinds of post-truth discourses. Furthermore, generative tools could produce a wave of content flooding cyberspace, which happens to be a very common access point to the world of knowledge. This content would in turn feed into the reference corpus, thereby creating a feedback loop with unpredictable consequences. However, the quality of the data produced by the mining that feeds the text generation process is a determining factor in the quality of its answers. In this respect, it is clear that unfiltered exploration of the internet \u2014 an immense web of information without hierarchy or labelling \u2014 constitutes a problem, despite the large number of parameters. The exploration of the noise of the world by automatons will become the exploration of a self-powered cacophony.<\/p>\n\n\n\n

This technology also calls into question the transmission and control of knowledge: ChatGPT has demonstrated its abilities in writing assignments ranging from school homework to doctoral theses, and academic institutions are buzzing with debate on the subject. For example, Sciences Po in France has opted for an outright ban<\/a>, while other institutions advocate teaching students to use it as a tool.<\/p>\n\n\n\n

Furthermore, the role of human actors in optimising system performance raises concerns about legitimacy and transparency. The process of moderation may introduce forms of censorship of all kinds. Might these poorly paid workers, employed to remove objectionable content from the AI\u2019s responses, end up rewriting history or deleting offensive content, such as certain satirical cartoons, even if the latter do not contravene the law? At the time of writing, OpenAI rejects searches concerning people and topics of a religious or sexual nature. But how long will this policy go on, and what will its competitors do?<\/p>\n\n\n\n

The attractions of ChatGPT are real and, at the same time, the proven risks are numerous. A number of them have been known since the beginning of the digital age, such as the spread of \u201cfake news\u201d, the dissemination of the biases present in training data, and the invasion of privacy and the cognitive dependency that come with the use of such tools.<\/p>\n\n\n\n

Once again, technological innovations are emerging at an accelerated pace. They are upsetting existing balances, and in this case our relationship to knowledge, its transmission, and its role in the social hierarchy. And we are only at the beginning. The growth of available data continues, and the calculating capacity of computers is increasing. There is no doubt that these systems will progress. A ChatGPT-4, with improved performance, has already been announced. These systems formally exploit human knowledge without understanding it, but they are capable of synthesising it and returning those responses to us. Admittedly, we are still far from a generalist AI,[3]<\/a><\/sup> but these applications will bring about major transformations, for which we must prepare ourselves.<\/p>\n\n\n\n

<\/em>This article has been translated from French by Sam Ferguson.<\/span><\/em><\/p>\n\n\n

    \n
  1. \n

    A test of an AI system\u2019s ability to mimic human conversation. \u2191<\/a><\/p>\n<\/li>\n

  2. \n

    BuzzFeed is an American news website and Internet media company. \u2191<\/a><\/p>\n<\/li>\n

  3. \n

    According to Andr\u00e9 Le Garff, author of the Dictionnaire de l\u2019informatique <\/em>(Paris: Presses universitaires de France, 1975): \u201cStrong or generalist AI refers to a machine capable not only of producing intelligent behaviour, including modelling abstract ideas, but also of experiencing a sense of real awareness and an understanding of its own reasoning\u201d. \u2191<\/a><\/p>\n<\/li>\n<\/ol><\/div>","protected":false},"excerpt":{"rendered":"

    The ability of the internet industry giants to bring about \u201cdisruptive\u201d innovations shows no sign of waning. On 30 November 2022, OpenAI \u2014 a company co-founded by Elon Musk and Sam Altman, and supported by Microsoft \u2014 announced the release of ChatGPT-3, the application that everyone is currently talking about. It is an artificial intelligence … Continued<\/a><\/p>\n","protected":false},"author":25,"featured_media":65487,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_seopress_titles_title":"","_seopress_titles_desc":"","_seopress_robots_index":"","_seopress_robots_follow":"","_seopress_robots_imageindex":"","_seopress_robots_snippet":"","_seopress_robots_primary_cat":"none","_seopress_robots_breadcrumbs":"","_seopress_robots_freeze_modified_date":"","_seopress_robots_custom_modified_date":"","_seopress_robots_canonical":"","_seopress_social_fb_title":"","_seopress_social_fb_desc":"","_seopress_social_fb_img":"","_seopress_social_fb_img_attachment_id":0,"_seopress_social_fb_img_width":0,"_seopress_social_fb_img_height":0,"_seopress_social_twitter_title":"","_seopress_social_twitter_desc":"","_seopress_social_twitter_img":"","_seopress_social_twitter_img_attachment_id":0,"_seopress_social_twitter_img_width":0,"_seopress_social_twitter_img_height":0,"_seopress_redirections_value":"","_seopress_redirections_enabled":"","_seopress_redirections_enabled_regex":"","_seopress_redirections_logged_status":"both","_seopress_redirections_param":"","_seopress_redirections_type":301,"_seopress_analysis_target_kw":"","footnotes":""},"categories":[140],"tags":[210,212],"keyword":[988,1239,1278],"class_list":["post-70135","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-newswatch","tag-science-et-technique-en","tag-societe-en","keyword-intelligence-artificielle-en","keyword-automatisation-en","keyword-technologie-de-l-information-en"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/posts\/70135","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/users\/25"}],"replies":[{"embeddable":true,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/comments?post=70135"}],"version-history":[{"count":5,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/posts\/70135\/revisions"}],"predecessor-version":[{"id":97160,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/posts\/70135\/revisions\/97160"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/media\/65487"}],"wp:attachment":[{"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/media?parent=70135"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/categories?post=70135"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/tags?post=70135"},{"taxonomy":"keyword","embeddable":true,"href":"https:\/\/www.futuribles.com\/en\/wp-json\/wp\/v2\/keyword?post=70135"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}