-
AIが要約したニュース、その半分は間違いだらけだと判明 | ギズモード・ジャパン
https://www.gizmodo.jp/2025/02/ai-is-not-good-at-summarizing-the-news.html2025.02.17 12:00
author R.Mitsubori書き手と読み手、どちらもAIに翻弄されないで。
イギリスの公共放送機関BBCが、AIが要約したニュース記事を調査したところ、その多くに｢重大な問題｣が含まれていることが判明しました。
この調査結果を受けて、BBCのデボラ・ターネスニュース担当最高責任者は、｢生成AIツールは火遊びをしている｣と警告し、国民の｢事実に対する脆弱な信頼｣を損なう恐れがあると指摘しています。
高確率でAIは要約を間違える
調査では、ChatGPT、Copilot、Gemini、Perplexityに100件のニュース記事を要約するよう依頼し、それぞれの回答を評価しました。すると、AI回答の51％に何らかの重大な問題があると判断されました。さらに、BBCのコンテンツを引用したAI回答の19％に、事実の記述、数字、日付の誤りなど、事実誤認があったそう。
-
人間が認識するのも誤認するからなぁ
ソース見れないし -
ソースを読まない嫌儲民が勝ってしまったか
-
ジャップならそもそもの官報ニュース自体が半分くらい役所の捏造だぞ舐めんなよ
-
AI「あびゃびゃびゃ」(いい加減なことを言う)
信者「これがシンギュラリティ…！」 -
>>6
晋ギュラリティ「私の要約は全く正しいと思いますよ、私は総理大臣なんですから 」 -
じゃあ朝日新聞のニュースは何割間違ってるの？
-
いずれ、AIがやってくれる事にケチをつけるな！となる
だから事実かどうかなんて無意味なんだよ -
半分は正しいことも言ってるのか？偉いじゃん進歩したな
-
2年後くらいには人間超えてるだろ
-
取り敢えず色んな生成AIをごった煮で評価を下してたりバージョン情報が無かったり怪しい記事だな
-
何も解決してないけど
そういうのはいいんだよ -
人間が書いて要約した記事は、8割が間違いだらけでしたってオチだろ。
-
原文見てもバージョンに関する情報がないね
ケチを付けたいという書き手の意思が伝わるね -
>>14
原文ってなにをさすかわかりませんが
こちらをよみました？
ps://www.bbc.co.uk/aboutthebbc/documents/bbc-research-into-ai-assistants.pdf -
そもそものＡＩが学習する情報源のインターネッツが間違いだらけなのでは？
名のある機関の出す情報は真実と言われそうでないものは陰謀や嘘と言われる
それをただ信じるだけの大衆
このひとつの考え方も陰謀や嘘なのかね？ -
>>17
名の有る機関は世に意見を出す前に複数のチェックをしてることが比較的多いからね
それだけ正確だ
たいして在野の独自研究は当人の思い込みだけで構成されてるから統合失調症の妄想となんにも変わらない
どちらが正しそうか、の判別がつかないとしたらお前の知的能力は著しく低いし
判別がつくのに人々に後者を信じさせようとするならお前はカルトの手先だ -
>>28
俺が言ったことをそのままなぞるこういう大衆のうちの一人がいる
つまりそういうことだ
大衆の多くは情報の真偽など問題としてないという証左ではなかろうか -
まぁgeminiがゴミなのはそうだけど
-
記者の最後の抵抗か
-
AIの成長するする詐欺やら驚き屋工作にも限界が来てるな
AIバブル崩壊は間近か -
A key feature of the emerging landscape is AI assistants like those from OpenAI, Google, and
Microsoft. AI assistants are adept at many tasks including drafting emails and documents;
analysing data; and summarising information. They can also provide answers to questions
about news and current affairs. They do this, in part, by repurposing content from publishers'
websites, often without publishers’ permission.
To better understand the news related output from AI assistants we undertook research into
four prominent, publicly available AI assistants – OpenAI’s ChatGPT; Microsoft’s Copilot;
Google’s Gemini; and Perplexity. We wanted to know whether they provided accurate
responses to questions about the news; and if their answers faithfully represented BBC news
stories used as sources.
We gave the AI assistants access to our website for the duration of the research and asked
them questions about the news, prompting them to use BBC News articles as sources where
possible. AI answers were reviewed by BBC journalists, all experts in the question topics, on
criteria including accuracy, impartiality and how they represented BBC content.
The BBC is the UK’s most widely used and trusted news provider and the world’s most trusted
international news provider.1,2 We take time and care to ensure the accuracy and impartiality
of our news. Our journalists approached this task with the same level of care. -
2
The answers produced by the AI assistants contained significant inaccuracies and distorted
content from the BBC. In particular:
• 51% of all AI answers to questions about the news were judged to have significant
issues of some form.
• 19% of AI answers which cited BBC content introduced factual errors – incorrect factual
statements, numbers and dates.
• 13% of the quotes sourced from BBC articles were either altered from the original
source or not present in the article cited.
This matters because it is essential that audiences can trust the news to be accurate, whether
on TV, radio, digital platforms, or via an AI assistant. It matters because society functions on a
shared understanding of facts, and inaccuracy and distortion can lead to real harm.
Inaccuracies from AI assistants can be easily amplified when shared on social networks. It
matters because news publishers must be able to ensure their content is being used with their
permission in ways that accurately represent their original content and reporting. We also
know from previous internal research that when AI assistants cite trusted brands like the BBC
as a source, audiences are more likely to trust the answer – even if it is incorrect.
Individual errors highlight some of the issues our research found. For example, Google’s
Gemini incorrectly stated that “The NHS advises people not to start vaping, and recommends
that smokers who want to quit should use other methods”. In fact, the NHS does recommend
vaping as a method to quit smoking. Microsoft’s Copilot incorrectly stated that Gisèle Pelicot
uncovered the crimes against her when she began having blackouts and memory loss. In fact,
she found out about the crimes when the police showed her videos they had found when they
confiscated her husband’s electronic devices. Perplexity mis-stated the date of Michael
Mosley’s death and misquoted a statement from Liam Payne’s family after his death. OpenAI’s
ChatGPT claimed in December 2024 that Ismail Haniyeh, who was assassinated in Iran in July
2024, was part of Hamas leadership.
Our research can only scratch the surface of the issue. The scale and scope of errors and the
distortion of trusted content is unknown. This is because AI assistants can provide answers on
a very broad range of questions and users can receive different answers to the same or similar
question. Audiences, media companies and regulators do not know the extent of the issue. It
may be that AI companies do not know either -
最近嫌儲にいるよやだよなAIでソース要約して掲載するゴミが
-
The implications of this research are far-reaching.
AI assistants cannot currently be relied upon to provide accurate news and they risk
misleading the audience. While AI Assistants often include a disclaimer about the risk of
inaccuracy there is no mechanism for AI applications to correct errors, unlike professional
news outlets that acknowledge and correct occasional errors.
It is likely other publishers may be similarly affected by the issues the research has identified.
It may also be that errors are repeated in other areas where reliability and accuracy is
paramount – for example, health, education and security.
We expect the use of AI assistants to grow so it's critical they provide audiences with accurate
and trustworthy information. Publishers, like the BBC, should have control over whether and
how their content is used and AI companies should show how assistants process news along
with the scale and scope of errors and inaccuracies they produce.
Ensuring people can find trusted information in the age of AI will require AI and media sectors
to work together, and the BBC is ready and willing to work closely with others. We are also
planning a series of AI literacy activities to help audiences navigate their use of AI. -
We believe there are three important next steps:
1) We want AI companies to hear our concerns and work constructively with us. We want
to understand how they will rectify the issues we have identified and discuss the right
long-term approach to ensuring accuracy and trustworthiness in AI assistants. We are
willing to work closely with them to do this.
2) Regulation may have a key role to play in helping ensure a healthy information eco-
system in the AI age. Policymakers are already looking at this area. For example Ofcom
has recently published research that shows that audiences that consume news from
Public Service Broadcasters (PSBs) have higher trust in institutions, better knowledge of
news and are less polarised. We believe it is vital that the accuracy of PSB news is
preserved through the prism of AI assistants. It is essential that PSBs, AI companies,
Ofcom and Government work together to secure an effective regulatory regime.
3) We will repeat this study in the near-future and believe there would be value in regular
evaluations which may sit best with a regulator or research institute – providing an
independent view of the accuracy and trustworthiness of news content on AI platforms,
including AI assistants.
The BBC is excited about the opportunities AI will bring. We look forward to working with the
AI and media sectors to ensure AI is used responsibly and brings value to our audiences.
Pete Archer, Programme Director Generative AI -
知ってた速報
ググると勝手にAIが答えてくるくせに、明らかに間違ってい情報を出してくるのはやめろ -
あいつら分からない事を分からないと言わないで捏造してくるからな
-
>>32
ほんとそれ
屁理屈でデタラメ言うなんてケンモメンかっつーの -
要約とかは進学校とかで訓練したりするもですが
日本人は結構苦手ですよね -
設定方法とか嘘八百並べられるから困る
-
AIで民意を組み上げて政策に反映させるとか言い出す流行り物好き馬鹿の典型の党首がいるらしいじゃないですか
-
情報源がクソなんだろ
-
うぜぇ短縮してもURL貼れねぇわ
記事の一番下の方にあるbbcのリンクな -
デマや根拠ない妄想をAIが判断出来ずにそのまま鵜呑みにしてしまうから間違った回答をしてしまう
コメント