Generative AI's Impact on Web Traffic and Decentralization: A New Challenge for Online Information Access

Geneva, Switzerland Switzerland
Centralized platforms like Facebook and Google have directed user traffic, eroding web decentralization.
China led the way in generative AI patent applications from 2014 to 2023 with over 38,000 inventions.
Generative AI can generate direct answers from aggregated data, reducing the need to visit individual websites.
Generative AI is making waves in technology with models like ChatGPT and Google Gemini.
Generative AI reduces web traffic, impacting revenue of original knowledge publishing websites.
The World Wide Web was created in 1989 and marked by decentralization and a wide array of content options.
Transformer models can be affected by tokenization issues leading to worse performance for users of less 'token-efficient' languages.
Generative AI's Impact on Web Traffic and Decentralization: A New Challenge for Online Information Access

Title: The Impact of Generative AI on Web Traffic and Decentralization

Generative artificial intelligence (AI) has been making waves in the technology world, with models like ChatGPT from OpenAI and Google Gemini generating text, images, music, and computer code. However, this advancement raises questions about its impact on web traffic and decentralization.

According to a report by the United Nations Intellectual Property Organization (UNIPO), China led the way in generative AI patent applications from 2014 to 2023, with over 38,000 inventions. This surge in generative AI development could potentially change how we access and consume information online.

Tom Wheeler, writing for Brookings Tech Tank, discusses the implications of generative AI on web traffic. He notes that these models can generate direct answers from aggregated data, reducing the need to visit individual websites. For instance, Apple's integration of OpenAI into Siri exemplifies this shift towards AI-based intermediaries.

However, this development could also lead to a managed decline in the web as we know it. The World Wide Web was created by Sir Tim Berners-Lee in 1989 and transformed the internet into a user-friendly network of diverse information sources. Berners-Lee emphasized that the first decade of the web was marked by decentralization and a wide array of content options.

But as centralized platforms like Facebook and Google have directed user traffic, generative AI is further eroding this vision by providing direct answers from aggregated data, potentially reducing website visits and even Google searches. This shift could impact commercial websites that rely on web traffic for revenue.

Moreover, the tokenization process used in transformer models can affect their performance. For example, some tokens have odd spacing which can derail a transformer and affect its performance. Transformers treat case differently and fail capital letter tests due to different tokenization methods for each character or word. Non-English languages that don't use spaces to separate words cause differences in tokenization methods and longer completion times for tasks in these languages, leading to worse model performance for users of less 'token-efficient' languages.

These issues could lead to a vicious cycle where generative AI reduces web traffic, impacting the revenue of original knowledge publishing websites. In turn, these websites may struggle to invest in research and development or even face closure. OpenAI charges users for using ChatGPT but pays nothing to millions of websites from where it collects knowledge.

Several class action lawsuits have been filed in the USA against companies for sucking up vast amounts of intellectual assets through generative AI without compensation. As we navigate this new landscape, it's crucial to consider the implications of generative AI on web traffic and decentralization.


Article Image
Article Image
Article Image
Article Image
Article Image
Article Image
Article Image
Article Image
Article Image
Article Image

Confidence

91%

Doubts
  • Is there a definitive number of patent applications for generative AI in China?
  • What percentage of web traffic is being redirected through centralized platforms versus directly through generative AI models?

Sources

95%

The overall score is a weighted number that takes into account conflict of interest, bias, deception and other practices that undermine the credibility of the source. It is calculated as:
(Site Conflicts Of Interest + Author Conflicts Of Interest) / 2.0 * 0.2 + ArticleBiasScore * 0.20 + UniquePointsScore * 0.05 + DeceptionScore * 0.20 + ReadabilityScore * 0.05 + FallacyScore * 0.20

A score that takes into consideration the content for flow, interruptions with ads, and overt search engine optimization techniques that makes the content hard to understand

Tokens are a big reason today's generative AI falls short

TechCrunch Kyle Wiggers Saturday, 06 July 2024 17:00
  • Unique Points
    • Generative AI models are built on an architecture known as the transformer.
    • , Transformers can’t take in or output raw text without a massive amount of compute.
    • , Text is broken down into smaller pieces called tokens during tokenization process.
    • , Some tokens have odd spacing which can derail a transformer and affect its performance.
    • , Transformers treat case differently and fail capital letter test due to different tokenization methods for each character or word.
    • , Many non-English languages don’t use spaces to separate words, causing differences in tokenization methods and longer completion times for tasks in these languages.
    • , Users of less ‘token-efficient’ languages are likely to see worse model performance and pay more for usage due to high token counts caused by different tokenization methods.
    • , Google DeepMind AI researcher Yennie Jun showed that some languages needed up to 10 times more tokens to capture the same meaning in English.
    • , Models struggle with math due to inconsistent digit tokenization, which destroys relationships between digits and results in transformer confusion.
  • Accuracy
    • Generative AI models are now able to generate direct answers from aggregated data, reducing web traffic and bypassing individual site visits.
    • Generative AI does not create original knowledge but compiles existing bits and pieces from numerous websites and recasts them as revenue-generating new products.
  • Deception (100%)
    None Found At Time Of Publication
  • Fallacies (100%)
    None Found At Time Of Publication
  • Bias (100%)
    None Found At Time Of Publication
  • Site Conflicts Of Interest (100%)
    None Found At Time Of Publication
  • Author Conflicts Of Interest (100%)
    None Found At Time Of Publication

96%

The overall score is a weighted number that takes into account conflict of interest, bias, deception and other practices that undermine the credibility of the source. It is calculated as:
(Site Conflicts Of Interest + Author Conflicts Of Interest) / 2.0 * 0.2 + ArticleBiasScore * 0.20 + UniquePointsScore * 0.05 + DeceptionScore * 0.20 + ReadabilityScore * 0.05 + FallacyScore * 0.20

A score that takes into consideration the content for flow, interruptions with ads, and overt search engine optimization techniques that makes the content hard to understand

Connecting The Dots: AI Is Eating The Web That Enabled It

PYMNTS.com Unknown CPI Friday, 05 July 2024 16:06
  • Unique Points
    • Tom Wheeler wrote an article for Brookings Tech Tank about the impact of generative AI on web traffic.
    • The World Wide Web, created by Sir Tim Berners-Lee in 1989, transformed the internet into a user-friendly network of diverse information sources.
    • Generative AI models are now able to generate direct answers from aggregated data, reducing web traffic and bypassing individual site visits.
    • Apple’s integration of OpenAI into Siri exemplifies this shift towards AI-based intermediaries that could potentially reduce website visits and Google searches for commercial websites.
  • Accuracy
    No Contradictions at Time Of Publication
  • Deception (80%)
    The article makes several statements that imply a negative impact of AI on the web and web publishers. While these statements are factually correct, they do not provide any new information or insights beyond what is already known. The author also uses emotional language to describe the situation, such as 'managed decline' and 'carnage'. These words are intended to manipulate the reader's emotions rather than providing objective analysis. Additionally, the article selectively reports on studies that support its position without mentioning any studies that contradict it.
    • Generative AI's ability to provide direct conclusions threatens the fundamental purpose of websites, especially those reliant on commercial support.
    • The web is entering a state of managed decline,
    • Web publishers brace for carnage as Google adds AI answers.
  • Fallacies (100%)
    None Found At Time Of Publication
  • Bias (100%)
    None Found At Time Of Publication
  • Site Conflicts Of Interest (100%)
    None Found At Time Of Publication
  • Author Conflicts Of Interest (100%)
    None Found At Time Of Publication

75%

The overall score is a weighted number that takes into account conflict of interest, bias, deception and other practices that undermine the credibility of the source. It is calculated as:
(Site Conflicts Of Interest + Author Conflicts Of Interest) / 2.0 * 0.2 + ArticleBiasScore * 0.20 + UniquePointsScore * 0.05 + DeceptionScore * 0.20 + ReadabilityScore * 0.05 + FallacyScore * 0.20

A score that takes into consideration the content for flow, interruptions with ads, and overt search engine optimization techniques that makes the content hard to understand

Has Generative AI triggered a vicious cycle?

The Financial Express M Rokonuzzaman Saturday, 06 July 2024 21:29
  • Unique Points
    • Generative artificial intelligence (Gen AI) does not create original knowledge but compiles existing bits and pieces from numerous websites and recasts them as revenue-generating new products.
    • Original knowledge publishing websites are suffering from decreasing visits and revenue due to the rise of Gen AI tools like ChatGPT.
    • OpenAI charges users for using ChatGPT but pays nothing to millions of websites from where it collects knowledge.
    • Several class action lawsuits have been filed in the USA against companies for sucking up vast amounts of intellectual assets through Gen AI without compensation.
  • Accuracy
    No Contradictions at Time Of Publication
  • Deception (35%)
    The author makes several assertions that border on editorializing and sensationalism. He implies that ChatGPT and other generative AI tools are 'stealing' knowledge from original publishers without consent, but he does not provide any evidence to support this claim. He also states that these tools are eliminating the need for users to visit websites, resulting in decreased revenue for publishers. However, he does not consider the possibility that users may prefer the convenience of accessing information through AI tools rather than manually searching for it on various websites. Additionally, he quotes a study by The Atlantic and a survey by the University of Toronto as evidence of this trend, but he does not disclose whether these sources have been peer-reviewed or not. He also makes several statements about the legal implications of AI tools collecting intellectual assets without consent, but he does not provide any specific examples or references to laws that have been broken. Overall, the author's article contains several instances of selective reporting and emotional manipulation.
    • To their surprise, that precious asset is being stolen, recast, and sold without even mentioning the creators’ names.
    • It’s worth noting that AdSense shares as high as 80 per cent of revenues with publishers.On the other hand, although OpenAI charges users for using ChatGPT, it pays nothing to millions of websites from where it collects knowledge.
    • ChatGPT, like Gen AI tools, generates revenue by selling compiled knowledge collected from original publishers without consent, which could be termed stealing.
    • For example, a survey by the University of Toronto reveals that 22 per cent of ChatGPT users ‘use it as an alternative to Google.’
    • Instead of visiting all those websites publishing original content, seekers prefer to ask questions to knowledge aggregators-Gen AI tools such as ChatGPT.
  • Fallacies (50%)
    The author makes several arguments that contain fallacies. Firstly, the author commits an appeal to ignorance fallacy when they claim that 'despite claims, no Gen AI has existed without human intelligence.' This statement is not supported by evidence and is based on the assumption that because human involvement is required in creating and training AI models, it cannot produce original knowledge. Secondly, the author uses a hasty generalization fallacy when they state 'For example, a survey by the University of Toronto reveals that 22 per cent of ChatGPT users "use it as an alternative to Google.' This statement is based on one survey and does not represent the entire user base of ChatGPT or other similar AI tools. Lastly, the author uses a dichotomous depiction fallacy when they present two extreme scenarios: either knowledge creators and publishers are thriving or they are being completely eradicated. The reality is likely somewhere in between.
    • ]Despite claims, no Gen AI has existed without human intelligence.[/...]
    • [For example, a survey by the University of Toronto reveals that 22 per cent of ChatGPT users "use it as an alternative to Google.']
    • [Instead of going back to the dark age, we need to draw lessons from the past about the benefits that the human race has already obtained by adopting policies and legal frameworks for offering economic incentives to millions of creative minds in generating, distributing, and pursuing knowledge and ideas in driving economic prosperity through finding better means in getting jobs done.]
  • Bias (100%)
    None Found At Time Of Publication
  • Site Conflicts Of Interest (100%)
    None Found At Time Of Publication
  • Author Conflicts Of Interest (100%)
    None Found At Time Of Publication

89%

The overall score is a weighted number that takes into account conflict of interest, bias, deception and other practices that undermine the credibility of the source. It is calculated as:
(Site Conflicts Of Interest + Author Conflicts Of Interest) / 2.0 * 0.2 + ArticleBiasScore * 0.20 + UniquePointsScore * 0.05 + DeceptionScore * 0.20 + ReadabilityScore * 0.05 + FallacyScore * 0.20

A score that takes into consideration the content for flow, interruptions with ads, and overt search engine optimization techniques that makes the content hard to understand

China is the runaway leader in generative AI patent applications followed by the US, the UN says

The Fixing Site: A Summary of the Article. Jamey Keaten | AP Wednesday, 03 July 2024 14:21
  • Unique Points
    • China requested more patents for generative AI than any other country with about 54,000 inventions from 2014 to 2023.
    • Over 38,200 generative AI inventions came from China, six times more than from the US.
    • Generative AI helps create text, images, music and computer code through tools like ChatGPT from OpenAI and Google Gemini.
  • Accuracy
    • Some critics fear that GenAI could replace workers in some jobs or improperly take human-generated content without fair compensation.
  • Deception (100%)
    None Found At Time Of Publication
  • Fallacies (95%)
    The article contains an appeal to authority with the statement 'China has requested far more patents than any other country when it comes to generative AI, the U.N. intellectual property agency said.' This is a fallacy because while the U.N. agency's statement is true, it does not necessarily imply that China leads in the development or implementation of generative AI technology.
    • China has requested far more patents than any other country when it comes to generative AI, the U.N. intellectual property agency said.
  • Bias (100%)
    None Found At Time Of Publication
  • Site Conflicts Of Interest (100%)
    None Found At Time Of Publication
  • Author Conflicts Of Interest (100%)
    None Found At Time Of Publication