Jin Lei from Au Fei Temple
Quantum Position | 公众号 QbitAI
What about the Intel ® Core ™ Ultra processor, which claims to be the biggest technological architecture change in 40 years?
Now, the most fashionable AI PCs have given a clear answer.
Without further ado, let's look at the effect.
For example, if you're going to immerse yourself in work for a while, you may need a laptop that meets the following requirements:
- Turn to silent
- Performance is adjusted to high-performance mode
- Open the file that you didn't finish reading yesterday
For such complex multiple requirements, now you don't need to use the mouse to set up and operate one by one, you only need to put forward requirements in a one-time way.
It's not just about working on multiple tasks and opening files quickly, it's about generating content summaries and ideas for files in a way that understands you.
For example, if you use an AI PC a lot, it will know that you like to summarize the content of the document first, and then do the work of ideas and logic.
And the longer you use the AI PC, the more it will understand how you work, and it will be able to give results the way you want and what you're used to without having to emphasize certain output points in the future.
What's more, all of this can be done without internet connection!
This is what Zhipu AI's PC intelligent assistant has been able to do on AI PCs equipped with Intel ® Core ™ Ultra processors.
Not only that, Zhang Fan, COO of Zhipu AI, also imagined that with the blessing of multimodal capabilities, AI PCs will bring more possibilities in the future.
For example, if you box select the content in an image, you can "snap" to get the details related to it.
Ask for a request and you can generate a poster in seconds.
Change the name of the file in batches, no need to manually change it one by one, just one sentence.
The daily tedious work of writing and sending emails can also be handled by AI PCs!
From this, it is not difficult to see the effect that Intel ® Core ™ Ultra processors have achieved so far -
Not only can it provide more powerful computing power support, but it can also play a "combination punch" with the popular large-scale model software applications, so that AIGC applications can run very efficiently and smoothly on commercial terminals.
However, the capabilities of PC smart assistants like the ones we just saw (i.e., AI PC assistants) are only a small part of what AI PCs with Intel ® Core ™ Ultra processors can do.
In fact, the computing power of this CPU has also been involved in more fields related to AI PCs, such as AI Office assistants, AI chatbots, AI local knowledge bases, AI video image processing, and AI PC management.
Next, let's continue to dig deeper into the performance of AI PCs with the blessing of Intel ® Core ™ Ultra.
Productivity Direct Pro Max
AI Office助手
If you want to ask the "three-piece suit" that everyone in the workplace can't avoid, it must be Word, Excel and PPT (what a painful understanding~~).
However, now with the blessing of AIGC, the "three-piece suit" may no longer be a headache for migrant workers.
Because products like ExtendOffice already have a new local-based AI plug-in function, it can be said that office efficiency can fly directly.
For example, with the cumbersome Excel, we can now solve all problems without rote memorization of complex formulas and functions, and only need natural language.
For example, if you say "USD to RMB" in Kutools AI, the value in the amount column will be changed automatically.
Encrypting and decrypting sensitive data can also be done in one sentence - "encrypt the selected cells".
In Word, it is still a sentence of "generate a marketing copy for an AI PC", and the content of the article is completed.
For example, for the highly repetitive and boring work of replying to work emails, everything can be generated under the AIGC capability.
Kutools AI can learn by itself based on historical emails, and then directly generate quote reply emails with precise numbers!
It is not difficult to see that in AI PCs, the original cumbersome operation of office software has now become a matter of one sentence.
AI Chatbot
In the matter of chat assistants, the gameplay of AI PCs has also begun to be different.
Taking Thundertech's legal model as an example, even on the end side of a laptop, it is possible to execute such tedious requirements as contract generation.
In addition to the entire contract generation, the legal model on the AI PC can also help lawyers and legal workers to do some case analysis, which can effectively help in basic judgment and preparation.
As mentioned above, since the large model of the AI PC can be run on the device side, even in scenarios such as airplanes, legal personnel who need to work can still run the large model locally to process contracts.
At the same time, the operation of the device side also means the security and privacy of data, which is exactly what legal personnel need when handling contracts.
AI local knowledge base
In AI PCs, RAG (Retrieval Enhanced Generation) is now also being given a new lease of life.
For example, the Boundless model brought by Transwarp is a large model that runs natively in an AI PC, making the experience very smooth at a speed of more than 20 characters per second; And it has enhanced its capabilities in pan-industry knowledge acquisition, content understanding and data analysis.
In addition, Transwarp also brought the RAG product of Boundless Ask, which can effectively convert massive local documents into knowledge with one click, and its features are as follows:
- Localized high-performance vector database that can run entirely locally on an AI PC.
- Support multiple formats, unlimited length, and automatic storage of knowledge and documents.
- Multimodal data such as images, audio, and video are supported.
- Self-developed RAG technology to achieve accurate Q&A.
- Enables accurate data analysis.
So how does it work?
Search for a piece of real-time information in Boundless Knowledge, it can not only interpret and answer the text content, but also make a knowledge graph of the key concepts involved, which can provide good help in daily work such as daily investment.
In terms of multimodality, Wuya Ask also supports image retrieval:
Even "feed" it a long video, and Boundless Ask can also make an accurate summary of it.
These two products have been released, and interested partners can give it a try~
AI multimedia processing
For the processing of videos, AI PCs also have a different way to play.
For example, for the slow motion of the video, Danghong Technology cooperated with Intel to launch a new solution on the AI PC, which can continue to insert 3 times the slow motion into the original 2x slow motion video.
Different from the traditional generation methods based on motion estimation and optical flow, more video frames can now be inserted through AIGC, and it is still smoother and better quality.
It's worth mentioning that this kind of work, which was previously only possible on professional servers and large workstations, can now be done with just an AI PC.
AI PC management
Although the applications brought by AIGC can be said to be "really fragrant", with the accelerated deployment of AI PCs, it will certainly bring problems to enterprises how to manage these new devices.
In this regard, in fact, many software vendors have already made efforts in this regard.
For example, velvet security applied AI to threat detection for the first time, and improved the overall efficiency of the anti-virus engine by running the AI model on the NPU and using OpenVINO for inference acceleration.
Another example is Berry Sunflower, which uses the powerful AI processing power of Intel NPU combined with OpenVINO+OCR technology to realize the recognition and real-time analysis of remote control images, and automatically generate a content backtracking report of remote control work.
VMware also integrates the remote management capabilities of the Intel ® vPro® platform with Intel ® Core ™ Ultra and provides an efficient and convenient security risk management solution under the operating system based on the device-side AI algorithm.
It is not difficult to see from various cases that AI PCs have begun to show their skills in a variety of different scenarios, and AI (especially large models) has also begun to release powerful processing power on the device side.
Then the next question is:
How?
On the device side, AIGC can show its strengths, and it must be inseparable from the support of strong computing power.
On the whole, we can attribute Intel's killer features in the AI PC era to: the "three AI engines" on the left hand and the "three laws" on the right.
The first is the three major AI engines, namely the CPU + GPU + NPU model adopted by Intel ® Core ™ Ultra, which makes it possible for 20 billion parameter large language models to run locally:
- CPU Engine: The central processing unit (CPU) in Intel ® Core ™ Ultra processors contains high-performance cores and energy-efficient cores that provide powerful data processing power and high data throughput, resulting in a significant increase in overall performance.
- GPU Engine: Intel Arc ™ graphics processing units (GPUs) offer approximately twice the performance gains and performance-per-watt optimization, which means faster processing speeds and greater energy efficiency when performing resource-intensive tasks such as video editing.
- NPU Engine: The Neural Network Processing Unit (NPU) is a hardware accelerator designed for AI computing, which is based on the advanced Intel 4 process technology and 3D high-performance hybrid architecture, supporting continuous AI application operation and greatly improving battery life.
The combination of these three AI engines enables AI PCs to efficiently run complex AI models locally, producing high-quality image and video content even without an internet connection.
This localized AI processing power not only optimizes the cost and efficiency of creative work, but also provides powerful support for commercial AI PCs.
As mentioned earlier, the Intel ® vPro® platform further combines the high performance of AI with the productivity, security, manageability, and stability required for business applications, enabling businesses to be more productive.
As for the three rules, they are summarized by Intel CEO Pat Gelsinger at the CES 2024 conference.
The first is the law of economics. This law emphasizes the importance of processing AI data locally to reduce reliance on cloud services and associated costs. By processing data directly on the device, cloud service costs can be reduced, making AI services more cost-effective.
The second is the laws of physics. The laws of physics focus on improving the speed and efficiency of data transfer between the cloud and on-premises devices. By optimizing network connectivity and hardware performance, AI services can respond quickly and solve bottlenecks that may arise during data transmission.
Finally, there is the law of data confidentiality. Data privacy laws focus on protecting the security and privacy of user data. This requires AI PCs to implement strict data protection measures during the design and implementation process to prevent data leakage and misuse and ensure the security of user information.
In Intel's view, the two-wheel drive between the "three engines" and the "three laws" is a solid foundation for the development of AI PCs, which can effectively promote the wide application and innovation of AIGC in various fields.
What does the popularity of AI PCs mean?
There is no doubt that the popularity of AI PCs will not be the kind of single-point explosion, but a phenomenon that will cause qualitative changes from multiple dimensions.
The most obvious thing is that for individuals, whether it is ExtendOffice changing the traditional "three-piece" gameplay of Office, or the mode of interaction with the PC has changed from personal operation to letting AI handle it by itself, it has become something that can be solved in one sentence.
As a result, users can significantly improve their productivity.
Secondly, if the six AI applications we just showed can be regarded as from point to line, then the full-stack AI solutions created by Intel and AI PCs will have an impact on the industry from line to surface.
For example, in the field of education, large language models running on AI PC terminals can significantly improve the efficiency and quality of teachers' lesson preparation, help solve the contradiction between teachers' work and learning, and promote the wide sharing of high-quality educational resources.
In the medical and pharmaceutical industries, models deployed on AI PCs can accelerate the discovery of new drug targets, thereby improving the efficiency of doctors and researchers in the drug development process.
In the field of rehabilitation and elderly care, AI PC evaluates the health status of the elderly through motion capture technology, and provides personalized rehabilitation and fitness guidance, using the power of science and technology to improve the health level of the elderly.
……
In short, the popularization of AI PCs will accelerate the realization of cost reduction and efficiency increase, and promote the digital transformation of various industries.
Finally, it's a new opportunity for Intel itself.
As the world's oldest chip giant, Intel has deep technology accumulation and market share in the field of PC hardware. A series of classics, such as 8088 processors, Pentium series processors, and Core series processors, have pushed Intel step by step to the one with the biggest PC hardware.
Now standing in the era of AI, although there are many "latecomers" who have seized the dividends of the times, Intel is also deeply integrating with AI with a strong momentum.
So will AI PCs make Intel greatest again? Time will tell.
— END —
量子位 QbitAI 头条号签约
Follow us and be the first to know about cutting-edge technology trends