この 1 年間、生成型 AI の開発は猛火に油が注がれて花が咲いたかのように進み、AI は時代の最も重要なトレンドになりました。今年はAIGCの技術から応用までの実践がキーワードとなる。したがって、AI PC は、当社の最も重要な制作ツールおよびコラボレーション ツールとして、間違いなく AI の理想を現実化する先駆者となっています。
AI PC が人間の生産性の解放を先導したいのであれば、明らかに何千もの家庭に普及するだけでは十分ではなく、何千もの産業に移行し、生産性の価値を真に凝縮する商業分野に利益をもたらす必要があります。
これに関連して、インテルは 3 月 26 日に「2024 New Intel Commercial Client AI PC Product Launch Conference」を開催し、AI PC が商用市場からどのようにスタートし、中国の産業データを加速するのかを確認することができました。知性の発達と変革。
このカンファレンスのインタビュー セッションでは、Intel クライアントがビジネスユニット副社長兼市場セグメント部門ゼネラルマネージャーの Feng Dawei 氏は、AI PC を一言で要約しました:
AI PC で最も重要なことは、コンピューティング アーキテクチャ。
この文章は、インテルがコンピューティング技術レベルで AI PC を構築するための基本的なロジックであると言えます。
AI の波を受けて、営利企業はシステムのセキュリティと管理性に対する要件が高まっていますが、同時に強力で低消費電力、インテリジェントな製品エクスペリエンスも求められており、新しい時代では企業にも求められています。製品のライフサイクル全体を通じてエネルギーを節約し、排出量を削減し、汚染を削減します。
インテル Core Ultra プロセッサーを搭載したインテル vPro プラットフォームは、基本的に、これらの変化に対応するために最下位レベルでコンピューティング アーキテクチャを再構築します。その中でも、当然ながら鍵となるのはインテル Core Ultra プロセッサーの AI 機能です。
Core Ultraプロセッサーの「再構築」、最も重要なのはCPU、GPU、NPUの3大AIエンジンの導入です。
CPU の強力なパフォーマンス コアとエネルギー効率コアは、より高いデータ スループットをもたらし、パフォーマンスを向上させます。
Intel Ruixuan GPU は、約 2 倍のパフォーマンスとパフォーマンス電力消費率をもたらし、ビデオ編集を高速化します。最新のソフトウェアのニーズを満たす AI やレイ トレーシングなどの高度な機能を備えています。
ローエンド向けに設計されたニューラル ネットワーク プロセッシング ユニット NPU 長時間実行される AI 向けに設計低消費電力でワークロードを処理できるため、前世代と比較して 2.5 倍のエネルギー効率を実現します。
さらに、NPU による継続的な AI アプリケーションのサポート、および Core Ultra によって採用されたインテル 4 プロセス テクノロジーと 3D 高性能ハイブリッド アーキテクチャ設計により、これらの機能により端末のバッテリー寿命が大幅に伸びる可能性があります。
これら 3 つの主要な AI エンジンはすべて AI 計算に使用でき、相互に連携してコンピューティング シナリオに基づいてどのコンピューティング ユニットを主に使用するかを決定するため、全体で最大 2.2 倍の世代向上をもたらします。 AIのパフォーマンス。
このような強力な AI パフォーマンスにより、Core Ultra は AI PC をサポートし、端末側がインターネットに接続されていないときに通常 200 億のパラメータを持つ大規模な言語モデルを実行し、高品質の写真やビデオを数秒で生成することができます。多くの AI アプリケーションは、デバイス上にローカルに配置するとスムーズに実行され、ユーザーにより効率的で便利なスマート プロダクション エクスペリエンスをもたらしますが、これまでは、複雑な AIGC 操作をサポートするために個人の PC でこのようなパフォーマンスを確認することは困難でした。
商用 AI PC の独自性に対応するだけでなく、インテル vPro プラットフォームは、AI と商用プラットフォームの生産性、セキュリティ、客観的合理性、安定性を有機的に組み合わせて、企業の効率化のニーズに応えます。安全性と安定性。
要約すると、Intel 中国技術部門のゼネラルマネージャーである Gao Yu 氏の言葉を借りれば、Intel Core Ultra プロセッサを搭載した商用 AI PC 製品には 3 つの大きな特徴があります。生産性パフォーマンスの大幅な向上、管理性の強化、そして 3 つの機能が組み込まれています。 -in AI エンジンは AI のパフォーマンスを飛躍的に向上させます。
インテルは、AI に最低レベルのコンピューティング能力 PC の商業利用は基礎を築いていますが、AI PC が商業分野で価値を完全に解放し、何千もの業界に力を与えることを促進するにはこれで十分だと思うなら、インテルはこう言うでしょう。これでは十分とは言えません。 。
商用分野で AI PC のより大きな可能性と価値を活用するために、インテルはさらに、経済学、物理学、データ機密性に関する 3 つの主要な法則と、次の 6 つの主要な AI シナリオを提案しました。すべての業界パートナーの焦点。
これは、商業分野での AI PC の実際の適用に基づいた彼らの考えと要約です。
具体的には、経済法則では、AI データのローカル処理によりクラウド サービスのコストが削減され、経済性が最適化されます。
物理法則では、データの適時性と正確性が求められます。 AI サービスでは、ボトルネックを解決するためにクラウドとエンド間の通信速度を向上させる必要があります。
In the data confidentiality rules, it is necessary to ensure the security and privacy of user data and prevent abuse and leakage.
In the past long period of industry exploration and discussion, the hybrid AI architecture of device and cloud collaboration has gradually become the most widely recognized basic architecture in the AI era. These three rules proposed by Intel, on the basis of device-cloud collaboration, point out the direction for AI PC to accelerate the transformation of creators' experience in the commercial field.
In other words, if you want to create a good commercial AI PC experience, these three rules are the foundation and must be followed.
For example, regarding the cooperation between end and cloud, Zhang Fan, chief operating officer of Beijing Zhipu Huazhang Technology Co., Ltd., who was an important ISV guest at this conference, talked about this issue. He said that from a device perspective, Intel Core Ultra is very strong, but from a model perspective, no amount of computing power may be enough.
Considering the particularity of commercial scenarios, many applications still want to be completed on the device side, so it is necessary to return to the device side more and maximize the capabilities of the device side. The ultimate, and then seamlessly integrate the capabilities of the cloud side with the client side, so that users can ensure high availability and stability while enjoying higher intelligence and a more convenient and secure experience.
It’s just that how to find a balance point that can satisfy users with a better experience based on the current terminal-side AI computing power still needs to be found in future practical applications.
Speaking of specific applications, Intel also provided six major AI scenarios that OEMs and ISV manufacturers need to focus on in this press conference, namely: AI Chatbot, AI PC Assistant, AI Office assistant, AI local knowledge base, AI image and video processing and AI PC management.
These six major AI scenarios may be the brightest sparks that we can see when AI and commercial PCs collide. AI serves thousands of industries. It will start from these six major AI scenarios, and then radiate to more specific and more segmented applications. In fact, in the exhibition area of this conference, we can already see some demonstrations for specific industry applications, such as The common goal that everyone in the commercial field desires: improved work efficiency and long battery life. These six major AI scenarios can really help everyone achieve this goal.
At the press conference, Intel worked with ISV partners to make specific demo demonstrations for each application scenario. Here, we might as well take the cooperation between Intel and Zhipu AI as an example to see how Core Ultra makes our ideal smart experience a reality.
Zhipu AI is a company transformed from the technological achievements of the Department of Computer Science of Tsinghua University, focusing on making large models. In January this year, they launched a new generation of large base model GLM-4, whose overall performance is comparable to the world's advanced level.
In the first demonstration, we can first see that if you want to know how to turn off the front camera of your computer, you only need to ask it in the AI PC smart assistant. will tell you the exact answer. Because many people are not familiar with the configuration and related operation methods of each computer, the electronic manual provided by the OEM manufacturer may not be easily found. The assistant developed by Zhipu is equivalent to using a large language model to convert the electronic manual The content is learned, and then it is used to answer people's questions about specific computer operations, which is undoubtedly more convenient and efficient.
The second scene demonstrates that when you are preparing to work, you can directly say to the smart assistant, "Please mute the computer volume first, switch to high-performance mode, and then open the My Yesterday A research case file that I haven’t finished reading.”
This is a relatively complex requirement, and the last file to be opened does not have a specific file name, which tests the user’s past operating habits of AI PC. understanding. We can see that the powerful AI capabilities of Core Ultra can indeed enable this AI PC to understand the presenter's needs in seconds, remember the user's past usage habits, complete all operations at once, and open files very accurately.
Then Zhipu AI demonstrated how to let the intelligent assistant summarize the content, opinions and development logic of the research case document just now according to the user's habits. Thanks to the powerful AI capabilities of the Core Ultra processor, this AI PC also completed the corresponding generation operations locally very quickly and accurately.
Moreover, Zhipu Zhang Fan said at the scene that as you use AI PC for longer and longer, it will learn more and more about your habits and way of thinking. You don’t even need to be specific about such functions, it will Able to refine and summarize documents according to your thinking habits.
Judging from the demonstration, intelligent assistants based on large end-side models such as Zhipu AI can run very smoothly and quickly on AI PCs equipped with Intel Core Ultra processors, and many complex and past tasks can be solved. We often need to perform functions on large cloud models, but now we can also have a good experience locally.
In order to achieve such an excellent client-side experience, Intel and Zhipu AI have also made a lot of systematic adaptation and optimization, ranging from response time to the number of tokens per second and other aspects. Experience indicators are considered so that the local large model of Zhipu AI and the performance call of Intel Core Ultra can achieve the most ideal and efficient cooperation state.
Of course, Zhipu also emphasized that the content of this demonstration is only a demo, not the final product, and the demonstration is currently only a large language model. During the exchange in the audience, this site also learned that Intel will continue to cooperate with Zhipu AI to gradually integrate the capabilities of multi-modal AI models into AI PCs with Core Ultra processors, including Wenshengtu, Diagram The ability to generate images, such as applications such as Image Generating, may also take advantage of the AI computing power of the cloud to bring a better experience through the cooperation of the device and the cloud.
This also reflects the balance and cooperation of device-cloud collaboration among the three principles we mentioned earlier.
Another good experience is that the AI experience shown in this demonstration will be integrated into AI PC products equipped with Intel Core Ultra processors as native solutions in the future and given to consumers. or customer. In other words, this will become a native capability of the Core Ultra AI PC. Whether it is a consumer or a commercial customer, no additional trouble is required. This is undoubtedly a very friendly thing, of course, provided that it is also Intel needs to cooperate with every OEM manufacturer.
In the interview, Intel China Technology General Manager Gao Yu also talked about this. He said that pre-installed bundling is definitely one of the models, and it will also be used in the early days. It's a faster way, but it's definitely not the only way. What the future business model
will look like still needs to be explored by the entire industry.
In any case, the cooperation between Intel and Zhipu AI still provides a "model experience" for commercial AI PCs, which allows the entire industry to see how Core Ultra AI PCs can be used in specific commercial applications. What kind of outstanding experience can be brought by using value in the scene, and what kind of endless imagination space can there be in the future.
As Zhang Fan said:
AI PC is a new paradigm of productivity. We should not overestimate its short-term effects, and at the same time, do not underestimate its long-term potential. It may bring us a new way of working that is completely different from the past, and also bring a new leap in productivity.
Not long ago, the market research organization IDC released the latest report, predicting that AI PC shipments will approach 50 million units in 2024, and will grow to 167 million units by 2027. , will account for about 60% of total global PC shipments.
It is worth mentioning that Intel launched the "AI PC Acceleration Program" last year, and one of its small goals is to implement artificial intelligence features on more than 100 million PCs by 2025. Therefore, IDC's data may be a "conservative estimate."
Not only that, on March 27, Intel also upgraded the "AI PC Acceleration Program", added the "AI PC Developer Program", and attracted independent hardware vendors (IHVs) to join "AI PC Acceleration Program."
In such a huge ecosystem, Intel’s role is, as repeatedly mentioned in this press conference:
“Water conservancy brings no dispute to all things” ."
I think that Intel and the industry to create full-stack AI solutions are "benefiting all things", and their efforts to implement and blossom commercial AI PCs and applications are "indisputable".
This is their "new productivity" and their "way" to promote the transformation of the digital and intelligent industry.
以上があなたの次のコンピューターは、信じられないほどの知性と心の知能を備えたものになるでしょう。の詳細内容です。詳細については、PHP 中国語 Web サイトの他の関連記事を参照してください。
![[Web フロントエンド] Node.js クイック スタート](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
