// “Pangu Chat” will primarily focus on political and enterprise applications, rather than being directly accessible to general consumers like ChatGPT. Analysts point out that Huawei has a more comprehensive industry chain and computing power allocation capability compared to other Chinese technology companies. When training the trillion-parameter Pangu model, Huawei utilized over 2,000 of its self-developed “Ascend 910” commercial AI chips. Huawei has also stated that the training cost for large-scale models over a span of three years amounted to a staggering 960 million Chinese yuan, indicating significant investment.

「盤古 Chat」將會主要針對政治和企業應用而設，並非像 ChatGPT 一樣直接向一般消費者開放。分析指華為比其他中國科技企業擁有更完整的產業鏈和算力調配能力，之前在訓練千億參數的盤古大模型時就採用了逾 2,000 個自家「昇腾 910」商用 AI 晶片，華為也曾表示 3 年間的大模型訓練成本高達 9.6 億人民幣，所費不菲 https://unwire.hk/2023/06/04/huawei-pangu-chat/fun-tech/?utm_source=rss&utm_medium=rss&utm_campaign=huawei-pangu-chat //

// OpenAI 表示，他們正在致力對抗「AI 幻覺」問題，嘗試用新的方式來訓練 AI 模型。這個而牛其實在生成式 AI 成為熱潮之前已經開始，發表的報告中承認，即使是最先進的 AI 模型，也一樣容易產生 AI 幻覺，在不確定內容的時候仍然會表現出「陳述事實」的傾向。一旦涉及推理的時候，這種邏輯錯誤就會令整個陳述存在問題。

OpenAI stated that they are actively working to combat the issue of “AI hallucinations” and are exploring new approaches to train AI models. They acknowledged in their published report that even the most advanced AI models are prone to generating AI hallucinations and have a tendency to make “factual” statements when faced with uncertain content. This logical fallacy becomes problematic when it comes to reasoning and can introduce flaws in the entire narrative https://unwire.pro/2023/06/01/openai-is-pursuing-a-new-way-to-fight-ai-hallucinations/ai/ //

// 根據兩位知情人士透露，MiniMax於2021年由包括前商湯科技（SenseTime）副總裁閆俊傑在內的一些前員工創立。

MiniMax was founded in 2021 by some former employees of SenseTime, including Yan Junjie – a former vice president at the Chinese AI firm, two other people said. https://www.reuters.com/technology/china-ai-startup-minimax-raising-over-250-mln-tencent-backed-entity-others-2023-06-01/ //

// James Cameron plans to incorporate the recent developments in artificial intelligence into the storyline of the next installment of “Terminator.” He believes that the previous film, “Terminator: Dark Fate,” didn’t perform as well as expected and hopes to revitalize the franchise by drawing inspiration from real-world advancements. Therefore, it is necessary to address the issues related to artificial intelligence before finalizing the script.

James Cameron 打算將近期的人工智能發展納入下一部《未來戰士》的故事情節中。他認為上一部的《未來戰士：黑暗宿命》表現未如理想，希望透過現實的啟發，令續作恢復活力，因此需要在完成最終劇本前釐清有關人工智能的問題。 https://unwire.hk/2023/06/01/chatgpt-james-cameron/fun-tech/ //

// 憑藉快速回應常見客戶詢問和基於對話分析推薦個性化產品或服務的能力，這個聊天機器人以企業上傳的知識庫作為唯一的信息來源，提供可靠準確的回答。

With the ability to quickly respond to common customer inquiries and recommend personalized products or services based on the conversation analysis, the chatbot provides reliable and accurate responses using businesses’ uploaded knowledge base as its sole information source https://finance.yahoo.com/news/hong-kong-startup-sleekflow-successfully-140000574.html //

// 這項技術的短期危險包括錯誤信息、隱藏的偏見和對邊緣群體和少數族裔的歧視。這些社會傷害是非營利研究機構（如由前谷歌人工智能研究員Timnit Gebru創辦的分散式人工智能研究所）的重點關注。

但像Bengio和Hinton這樣的人也擔心長期威脅，這並不是指今天的聊天機器人——它們能夠像人一樣回答問題。相反，他們擔心研究可能會迅速進展，使得在本十年內人工智能系統更加自主。

The near-term dangers of the technology include misinformation, insidious bias and discrimination against marginalised groups and minorities. These societal harms are the focus of non-profit research organisations such as the Distributed AI Research Institute, which was founded by former Google AI researcher Timnit Gebru.

But the long-term threats that also worry the likes of Bengio and Hinton do not refer to today’s chatbots — which provide human-like answers to questions. Instead, they fear that research could progress quickly to make AI systems more autonomous within this decade. https://www.ft.com/content/b4baa678-b389-4acf-9438-24ccbcd4f201 //

// Sam Altman’s 3-points plan:

Form a new government agency charged with licensing large AI models, and empower it to revoke that license for companies whose models don’t comply with government standards.

Create a set of safety standards for AI models, including evaluations of their dangerous capabilities. For instance, models would have to pass certain tests for safety, such as whether they could “self-replicate” and “exfiltrate into the wild” — that is, to go rogue and start acting on their own.

Require independent audits, by independent experts, of the models’ performance on various metrics. https://unwire.hk/2023/05/18/ai-15/fun-tech/?utm_source=rss&utm_medium=rss&utm_campaign=ai-15 //

// Caryn Marjorie, a 23-year-old influencer with 1.8 million followers on Snapchat, has utilized OpenAI’s GPT-4API technology and combined it with previously deleted video materials from her YouTube channel. After investing over 2,000 hours in design and development, she created “CarynAI.” The aim is to allow users to have direct conversations with Caryn herself using AI, creating an experience akin to being in a relationship with her.

Snapchat 上擁有180萬追蹤者的23歲 KOL Caryn Marjorie，利用 OpenAI 開發的 GPT-4API 技術，並結合曾在 YouTube上 發布過但已經刪除的影像素材，其後花費2000多個小時設計開發出「CarynAI」，希望用戶能利用 AI 直接和 Caryn 本人交談，就像與其戀愛一樣 https://unwire.hk/2023/05/12/kol-ai/fun-tech/?utm_source=rss&utm_medium=rss&utm_campaign=kol-ai //

// Mustafa Suleyman recently stated, “There is no doubt that many white-collar jobs will be very different in the next five to ten years, and there will be a lot of losers.” He believes that those who lose their jobs will become dissatisfied and radicalized, so governments need to consider how to support these unemployed individuals. One possible solution is a universal basic income, which addresses the livelihood issue of the unemployed population.

Mustafa Suleyman 最近表示：「毫無疑問，未來五到十年，白領階層的許多工作將變得非常不同，將會出現大量的失敗者。」他認為，這些失去工作的人們將會變得不滿且激進，因此政府需要考慮如何支援這些失業者，其中一個方式就是全民基本收入，解決失業人口的生計問題 https://unwire.pro/2023/05/11/ai-will-create-a-serious-number-of-losers-warns-deepmind-co-founder/ai/ //