generative AI Archives - Counterpoint

MediaTek Strengthens Premium Push With Gen AI Capabilities

The Dimensity 9300 is integrated with the 7th-gen APU 790 processor, which supports a broad generative AI ecosystem and models including Meta Llama 2, Baidu ERNIE 3.5 SE and Baichuan 2.
It can run with 13 billion parameters, which are scalable up to 33 billion, on-device at a processing speed of up to 20 tokens per second.
The Dimensity 9300 will debut with the vivo X100 smartphone in mid-November

MediaTek recently launched its third-generation premium chipset Dimensity 9300. The company has emerged as a strong competitor in the premium smartphone chipset market after launching the Dimensity 9000 in Q1 2022. Currently, MediaTek leads the low-mid segment and drives significant volumes both for 4G and 5G in this tier.
Adopting a non-traditional CPU core design, the Dimensity 9300 focuses on raw performance. It has four large cores (Cortex-X4) and four performance cores (Cortex-A720 cores), thus enabling it to excel in raw computing power and advanced AI capabilities.

Dimensity 9300 key specifications

Like the last two generations, the Dimensity 9300 SoC is built on a TSMC 4nm process node. It is more efficient and performs better than the Dimensity 9200. There is a 40% improvement in the multi-core performance and a 15% improvement in the single-core performance. The Dimensity 9300 combines an octa-core CPU with the company’s second-generation hardware raytracing engine, enabling smartphones to achieve console-level global illumination effects at a smooth 60 FPS. Besides, the chipset supports seamless multitasking, allowing users to simultaneously play games and stream videos or watch a video while gaming.

Four ARM Cortex-X4 CPU. Prime core clocked at up to 3.25GHz
Four ARM Cortex-A720 CPU clocked at up to 2.0GHz

7th-gen APU 790 processor

The chip is equipped with MediaTek’s next-generation APU 790 processor, which reduces power consumption by 45% while improving performance. Its processing speed is eight times that of the APU 690. It also offers significant improvements in generative AI performance and energy efficiency for edge computing. The APU 790 is specifically designed for generative AI tasks, marking a substantial upgrade over its predecessor. It accelerates processing through the Transformer model and supports image generation within one second using Stable Diffusion. The APU 790 also supports large language models with up to 33 billion parameters. MediaTek has also implemented mixed-precision INT4 quantization technology and NeuroPilot memory hardware compression to optimize memory usage for large AI models.

The Dimensity 9300 has a strong AI generative ecosystem, which supports language models like Llama 2, Baichuan 2 and Baidu AI LLM. It helps developers to efficiently deploy multi-modal generative AI applications for users.

Immortalis-G720 GPU

With the integration of ARM’s latest GPU, the Immortalis-G720, the Dimensity 9300 offers almost a 46% boost in GPU performance and 40% power reduction compared to the Dimensity 9200.

The Dimensity 9300 chipset supports the new Ultra HDR format in Android 14, improving mobile photography with vibrant images and compatible JPEG files. It also offers ambient light adaptive HDR recovery technology for enhanced photography. It supports 100% pixel-level autofocus, dual lossless zoom and 3-microphone HDR audio recording.

The chipset’s display system is equipped with the MiraVision Picture Quality (PQ) engine which dynamically adjusts the contrast, sharpness and color of primary objects, resulting in lifelike video experiences similar to high-end TVs. It uses on-device AI to detect primary objects and background images in real time.

Enhanced connectivity

The Dimensity 9300 offers Wi-Fi 7 speeds up to 6.5 Gbps and improved long-range connectivity with Xtra Range 2.0 Technology. It also enhances smartphone tethering speeds by up to three times using Multi-Link Hotspot technology. The Dimensity 9300 also supports up to three Bluetooth antennas and features dual Bluetooth flash connection technology for an ultra-low latency Bluetooth audio experience.

DRAM support and security

The Dimensity 9300 is the first SoC that supports the LPDDR5T up to 9600 Mbps. Also, it integrates two SUPs, one for boot security and one for computing security.

Dimensity 9300 vs Snapdragon 8 Gen 3

In terms of specifications, the Dimensity 9300 uses all big core architecture 4 prime cores (Cortex-X4) and 4 big cores (Cortex-A720), whereas the Snapdragon 8 Gen 3 uses one prime (Cortex-X4), five big (Cortex-A720) and two small cores (Cortex-A520). MediaTek with its all-big core design is addressing generative AI and gaming applications. On paper, the Dimensity 9300’s AI performance is competitive. The Dimensity 9300 supports large language models that can run with 13 billion parameters, whereas the Snapdragon 8 Gen 3 can run with 10 billion parameters on-device.

The fact that MediaTek now offers performance and efficiency gains that are comparable to Qualcomm’s latest-generation flagship offerings, shows MediaTek wants to directly compete with Qualcomm in the premium segment. Overall, this is going to be a win-win for the industry, as it will raise the bar and, in turn, benefit the end users.

Expected timeline

The vivo X100 will be the first smartphone to carry the Dimensity 9300 chipset. It will be available in the market by the end of 2023. We expect that the Dimensity 9300 will have better adoption among Chinese OEMs compared to the Dimensity 9200. China will be the first target market for smartphones with the Dimensity 9300, followed by India, SEA and Europe.

Qualcomm Snapdragon 8 Gen 3 Unveiled: On-Device Generative AI Takes Center Stage

On-device generative AI was a key theme of the launch.
Qualcomm has revamped AI Engine with support for LLM, LVM, and ASR to run solely on-device.
It can run 10 billion parameters on-device and LLM models at up to 20 tokens per second.

Qualcomm announced the next-generation Snapdragon 8 Gen 3 mobile platform at its annual Snapdragon Summit held on October 24-26 in Hawaii. One of the key highlights of the new system-on-a-chip (SoC) is that it is now capable of running accelerated AI computer engines on-device. The chipset extends the possibilities of generative AI and enables a rich user experience. The launch will strengthen Qualcomm’s lead in the hardware-based AI capabilities for mobile devices.

On-device Generative AI: The Key Theme

Qualcomm spent a lot of time talking about generative AI and how it will transform smartphone computing and user experiences. Last year, with the Snapdragon 8 Gen 2, the focus was on the Hexagon processor with Sensing Hub and on-device personalization. This year’s focus was on running Large Language Models (LLM), Language Vision Models (LVM), and transformer network-based automatic speech recognition (ASR) with up to 10 billion parameters running natively on-device.

Qualcomm has refreshed the AI engine with the Hexagon NPU showing double the performance growth and a 40% increase in the performance per watt to run the AI models. The chipset can now run up to 10 billion parameters on-device and LLM models at up to 20 tokens per second, removing the need to rely on the cloud for inferencing. Qualcomm has partnered with Meta to support Llama 2 and with Microsoft for Stable Diffusion. The Snapdragon 8 Gen 3, using Stable Diffusion on-device, can generate images in less than a second, without connecting to the internet. These are impressive capabilities for a low-power battery-operated device.

Source: Qualcomm

WATCH: On-Device Generative AI: Text-to-Image in Under a Second

Qualcomm has also added support on the SDK level so that developers can include their own models. Xiaomi has added a model for running AI on-device with six billion parameters. It has also added a host of support for creators to use multi-modal generative AI like:

LLM-based assistants can help the creative process and summarize ideas for you.
LVM can bring these ideas to life with the use of fast diffusion technology. For instance, you can ask to generate a completely new image, say, a family of four on the beach eating burgers.

Cognitive AI, i.e. LLM and LVM together, give the best user experience. HONOR announced that the Magic6 series smartphones will be powered by the Snapdragon 8 Gen 3 SoC and will have on-device LLM with seven billion parameters.

Source: Qualcomm

These models are being used for a breakthrough in camera experiences like:

A video object eraser by Arcsoft, allows to remove unwanted objects, and video bombers from videos.
OEMs can even use two always-sensing cameras in the front and back enabling easy QR code scan and face unlock.
The neural network allows you to zoom out beyond the picture captured, thus adding missing parts to photos making it feel like it was captured by a wide-angle lens. Qualcomm demoed this in action, calling it Photo Expansion, and it looked promising.

With advancements in AI-generated content, Qualcomm has also partnered with Truepic to adopt the C2PA standard. It lets viewers know whether the image is genuine or AI-generated and uses Qualcomm mobile security to create a cryptographic seal that indicates a genuine image. As all these experiences are on-device, the data never leaves your device, thus increasing security.

Source: Qualcomm

Snapdragon 8 Gen 3: Key Specifications

The Snapdragon 8 Gen 3 SoC is built on the enhanced TSMC 4nm process node, just like the last two generations. It is impressive to see how Qualcomm still managed to increase the CPU performance by making it 30% faster, while also consuming 20% less power. This was achieved by using a tri-cluster eight-core CPU design featuring:

One ARM Cortex-X4–based prime core clocked at up to 3.3GHz
Five ARM Cortex-720-based performance cores clocked between 3.0-3.2GHz
Two ARM Cortex-520-based efficiency cores clocked at up to 2.3GHz

In terms of graphics, Qualcomm claims that the new Adreno GPU is 25% faster in performance and 25% more power efficient. Overall power savings have improved by 10%, Qualcomm said.

Source: Qualcomm

Qualcomm has also improved the hardware-based ray tracing by 40% and added support for Unreal Engine 5 Lumen. The new Snapdragon 8 Gen 3 is the first chipset to support this engine. There is also the Adreno Frame Motion Engine 2.0, which doubles the frame rate from 60fps to 120fps. Furthermore, with the Snapdragon Elite Gaming suite, the new SoC also supports 240fps gaming on a 240Hz display.

Source: Qualcomm

Enhanced Connectivity: 5G Advanced Ready, Wi-Fi 7 and Dual Bluetooth

Qualcomm has added support for 5G advance using the Snapdragon X75 Modem-RF System, with hardware-based AI acceleration, making it another first for Qualcomm. This enables the Snapdragon 8 Gen 3 chipset to achieve better speeds, coverage, mobility, link robustness, and location accuracy. There is also the Qualcomm FastConnect 7800 Wi-Fi 7 platform that supports the High Band Simultaneous Multi-Link for faster speeds and low-latency performance.

OEM Partners and Expected Timeline

The Snapdragon 8 Gen 3 chipset has secured design wins among global OEMs, including ASUS, HONOR, iQOO, MEIZU, NIO, Nubia, OnePlus, OPPO, realme, Redmi, RedMagic, Sony, vivo, Xiaomi, and ZTE.

This is the first time that the launch of the chipset and device occurred simultaneously. Xiaomi took the limelight by launching the Xiaomi 14 series powered by the Snapdragon 8 Gen 3 SoC at the same time as the chipset launch. The Snapdragon 8 Gen 3 shipments will ramp up in Q1 2024 with wider adoption across the smartphone OEMs.

Qualcomm S7, S7 Pro Gen 1 Sound Platforms with AI-enhanced Audio and XPAN Technology

Along with the mobile platform, Qualcomm also announced the S7 and S7 Pro Gen 1 sound platforms that aim to offer a more advanced and personalized audio experience. Just like in the Snapdragon 8 Gen 3 mobile platform, AI follows a similar trajectory here as well. The sound platforms come with a dedicated AI coreto offer up to 100x higher AI performance, and five times more computing power compared to the previous generation.

Source: Qualcomm

The dedicated AI cores are also used for audio curation including hearing loss compensation, low-latency DSP and to power the latest fourth-generation Adaptive ANC. The platforms support Bluetooth 5.4, and Bluetooth LE including Auracast Broadcast Audio.

The Qualcomm S7 Pro Gen 1 platform takes it up a notch by adding micro-power Wi-Fi connectivity to intelligently switch between Bluetooth and Wi-Fi to offer the whole home and building coverage with XPAN technology. It supports 2.4GHz, 5GHz, and even 6GHz Wi-Fi bands. Lastly, with Snapdragon Sound, the S7 Pro Gen 1 can offer data rates of up to 29Mpbs, enabling lossless music streaming over Wi-Fi up to 24-bit/192kHz.

Snapdragon Seamless: Enabling Interoperability Between Windows and Android Devices

The Android and Windows ecosystems are fragmented, which makes interoperability a challenge. In an attempt to make devices work better together, irrespective of the OEM, Qualcomm also announced Snapdragon Seamless, which will enable Windows PCs and Android devices to discover each other and communicate seamlessly.

Source: Qualcomm

Samsung, with its Galaxy ecosystem, is trying to offer a seamless connectivity experience, but that is restricted to its own devices. Qualcomm is trying to break that barrier by offering a multi-device, multi-connect experience. Snapdragon Seamless will unify all devices from TWS and laptops to PCs, tablets, and smartphones, allowing rich experiences such as earbuds switching, text copy/paste, and image and video drag and drop between devices. It eliminates the need to constantly pair and unpair devices. Qualcomm demoed this between HONOR devices, and it looks promising.

Guest Post: OpenAI Haymaker?

OpenAI makes hay while the sun shines.

When a company that has issues with making profits can raise money at a valuation of $85 billion, it becomes abundantly clear that investors in generative AI have taken leave of their senses.

Open AI is reportedly raising money at a valuation of $80 billion to $90 billion. This looks like an opportunistic event for two reasons.

First, doubts over whether Open AI actually needs the money. It was only nine months ago that Microsoft invested $10 billion in OpenAI, meaning that if it has run out of money already, then it has a cash burn of $1.1 billion per month. This is Reality Labs’ levels of cash burn which with 400 employees amounts to $2.75 million per employee per month.

The vast majority of this spend will be going to compute costs where even with 100 million users making 30 requests per day this is an uneconomic level of spending. This would mean that ChatGPT and generative AI generally can never become a viable business or generate a positive ROI and so one suspects that OpenAI has in fact got plenty of money left.

Second, virtually free money. In the market’s mind, OpenAI is the leading generative AI company in the world (which is debatable). Furthermore, generative AI is the hottest theme in the technology sector by a wide margin, meaning that OpenAI sits at the pinnacle of what the market wants to own. This in turn means that OpenAI can sell far fewer shares for the money it wants to raise, and its existing shareholders can also register large unrealized gains on their balance sheets. Consequently, I think that this raise is opportunistic in that the market has given OpenAI an opportunity to capitalize on its fame and popularity.

However, most telling of all is that employees will also have an opportunity to sell some of their shares as part of this transaction. Insider stock sales are often an indicator of the insiders’ view that the valuation of the shares has hit a peak. At $85 billion, this is pretty hard to argue against.

OpenAI is supposed to earn revenues of $250 million this year and $1 billion next year, putting the shares on over 80x 2024 revenues. This is very high even in the best of times, but the plethora of start-ups and the thousands of models being made available for free by the open-source community leads one to think that competition is on the way.

Hence, price erosion is likely which in turn could lead to OpenAI missing the $1-billion revenue estimate for 2024 and burning through even more cash than expected. OpenAI will not be alone, and many start-ups will suffer from price erosion that will cause their targets to be missed. This could well be the pin that pricks the current bubble, causing enthusiasm to wane and valuations to fall.

OpenAI may not be worth $85 billion but the timing of the raise is perfect.

(This guest post was written by Richard Windsor, our Research Director at Large. This first appeared on Radio Free Mobile. All views expressed are Richard’s own.)

Podcast #70: Qualcomm Driving On-device Generative AI to Power Intelligent Experiences at the Edge

Generative AI like ChatGPT and Google’s Bard have disrupted the industry. However, they are still limited to browser windows and smartphone apps, where the processing is done through cloud computing. That is about to change soon as Qualcomm Snapdragon-powered devices will soon be able to run on-device generative AI.

At MWC 2023, Qualcomm showcased Stable Diffusion on a Snapdragon 8 Gen 2-powered Android smartphone. The demo showed how a smartphone can generate a new image with text commands or even change the background, without connecting to the internet. Running generative AI apps directly on a device offers several advantages, including lower operational costs, better privacy, security, and reliability of working without internet connectivity.

ALSO LISTEN: Podcast #69: ChatGPT and Generative AI: Differences, Ecosystem, Challenges, Opportunities

In the latest episode of ‘The Counterpoint Podcast’, host Peter Richardson is joined by Qualcomm’s Senior Vice President of Product Management Ziad Asghar to talk about on-device generative AI. The discussion covers a range of topics from day-to-day use cases to scaling issues for computing resources and working with partners and the community to unlock new generative AI experiences across the Snapdragon product line.

Click the play button to listen to the podcast

You can read the transcript here.

Podcast Chapter Markers

01:35: Ziad starts by defining generative AI and comparing it with machine learning and other types of AI.

03:56: Ziad talks about AI experiences that are already present in Snapdragon-powered devices.

06:24: Ziad addresses the scaling issue for computing resources used to train large language models.

09:46: Ziad deep dives into the types of day-to-day applications for generative AI on devices like a smartphone.

13:34: Ziad talks about the hybrid AI model, involving both cloud interaction and edge.

15:43: Ziad on how Qualcomm is leveraging its silicon chip capabilities to unlock generative AI experiences.

19:20: Ziad on how Qualcomm is working with its ecosystem and the developer community.

21:57: Ziad touches on the privacy and security aspect with respect to on-device generative AI.

Also available for listening/download on:

Podcast #69: ChatGPT and Generative AI: Differences, Ecosystem, Challenges, Opportunities

Generative AI has been a hot topic, especially after the launch of ChatGPT by OpenAI. It has even exceeded Metaverse in popularity. From top tech firms like Google, Microsoft and Adobe to chipmakers like Qualcomm, Intel, and NVIDIA, all are integrating generative AI models in their products and services. So, why is generative AI attracting interest from all these companies?

While generative AI and ChatGPT are both used for generating content, what are the key differences between them? The content generated can include solutions to problems, essays, email or resume templates, or a short summary of a big report to name a few. But it also poses certain challenges like training complexity, bias, deep fakes, intellectual property rights, and so on.

In the latest episode of ‘The Counterpoint Podcast’, host Maurice Klaehne is joined by Counterpoint Associate Director Mohit Agrawal and Senior Analyst Akshara Bassi to talk about generative AI. The discussion covers topics including the ecosystem, companies that are active in the generative AI space, challenges, infrastructure, and hardware. It also focuses on emerging opportunities and how the ecosystem could evolve going forward.

Click to listen to the podcast

Click here to read the podcast transcript.

Podcast Chapter Markers

01:37 – Akshara on what is generative AI.

03:26 – Mohit on differences between ChatGPT and generative AI.

04:56 – Mohit talks about the issue of bias and companies working on generative AI right now.

07:43 – Akshara on the generative AI ecosystem.

11:36 – Akshara on what Chinese companies are doing in the AI space.

13:41 – Mohit on the challenges associated with generative AI.

17:32 – Akshara on the AI infrastructure and hardware being used.

22:07 – Mohit on chipset players and what they are actively doing in the AI space.

24:31 – Akshara on how the ecosystem could evolve going forward.

Also available for listening/download on:

Guest post: AI Business Model on Shaky Ground

OpenAI, Midjourney and Microsoft have set the bar for chargeable generative AI services with ChatGPT (GPT-4) and Midjourney costing $20 per month and Microsoft charging $30 per month for Copilot. The $20-per-month benchmark set by these early movers is also being used by generative AI start-ups to raise money at ludicrous valuations from investors hit by the current AI FOMO craze. But I suspect the reality is that it will end up being more like $20 a year.

To be fair, if one can charge $20 per month, have 6 million or more users, and run inference on NVIDIA’s latest hardware, then a lot of money can be made. If one then moves inference from the cloud to the end device, even more is possible as the cost of compute for inference will be transferred to the user. Furthermore, this is a better solution for data security and privacy as the user’s data in the form of requests and prompt priming will remain on the device and not transferred to the public cloud. This is why it can be concluded that for services that run at scale and for the enterprise, almost all generative AI inference will be run on the user’s hardware, be it a smartphone, PC or a private cloud.

Consequently, assuming that there is no price erosion and endless demand, the business cases being touted to raise money certainly hold water. While the demand is likely to be very strong, I am more concerned with price erosion. This is because outside of money to rent compute, there are not many barriers to entry and Meta Platforms has already removed the only real obstacle to everyone piling in.

The starting point for a generative AI service is a foundation model which is then tweaked and trained by humans to create the service desired. However, foundation models are difficult and expensive to design and cost a lot of money to train in terms of compute power. Up until March this year, there were no trained foundation models widely available, but that changed when Meta Platforms’ family of LlaMa models “leaked” online. Now it has become the gold standard for any hobbyist, tinkerer or start-up looking for a cheap way to get going.

Foundation models are difficult to switch out, which means that Meta Platforms now controls an AI standard in its own right, similar to the way OpenAI controls ChatGPT. However, the fact that it is freely available online has meant that any number of AI services for generating text or images are now freely available without any of the constraints or costs being applied to the larger models.

Furthermore, some of the other better-known start-ups such as Anthropic are making their best services available online for free. Claude 2 is arguably better than OpenAI’s paid ChatGPT service and so it is not impossible that many people notice and start to switch.

Another problem with generative AI services is that outside of foundation models, there are almost no switching costs to move from one service to another. The net result of this is that freely available models from the open-source community combined with start-ups, which need to get volume for their newly launched services, are going to start eroding the price of the services. This is likely to be followed by a race to the bottom, meaning that the real price ends up being more like $20 per year rather than $20 per month. It is at this point that the FOMO is likely to come unstuck as start-ups and generative AI companies will start missing their targets, leading to down rounds, falling valuations, and so on.

There are plenty of real-world use cases for generative AI, meaning that it is not the fundamentals that are likely to crack but merely the hype and excitement that surrounds them. This is precisely what has happened to the Metaverse where very little has changed in terms of developments or progress over the last 12 months, but now no one seems to care about it.

(This guest post was written by Richard Windsor, our Research Director at Large. This first appeared on Radio Free Mobile. All views expressed are Richard’s own.)

AI Voice Assistants to Push Success of Autonomous Driving, Software-defined Vehicle

AI voice assistants are being integrated into cars for hands-free and intuitive functionality.
Voice assistants like Google Assistant and Apple Siri can recognize and respond to natural language commands, allowing drivers to interact with their vehicles more effectively.
Integrating natural voice virtual assistants is complex and requires significant resources and expertise in learning and data collection. As a result, only a few companies can currently do it successfully.

ChatGPT’s popularity has encouraged many people to think about AI’s potential applications. One of them is in the automotive sector. With the simplification of the dashboard in vehicles, there has been a trend towards integrating more functions into the central display, such as navigation, entertainment, climate control and vehicle diagnostics. The central computer in vehicles is becoming more powerful and can do more things. All this allows easier and more user-friendly ways for drivers to interact with their vehicles while enabling more advanced and customizable functions for the vehicle itself.

Also, this has matched the development of software-defined vehicles, which take this integration a step further by using a centralized software architecture to control all vehicle functions. This allows for greater flexibility and the ability to update vehicle systems over the air (OTA).

There has been an increasing demand for additional functions to be integrated into the central display, such as voice assistant, in-car digital assistant, and other advanced driver assistance systems (ADAS). However, oversimplification leads to many problems. Some people still like to use knobs or buttons in the auto cabin, despite the prevalence of touchscreen displays in modern cars. Below are some reasons:

Tactile feedback: Many people find it more intuitive to use these physical controls than to navigate through a digital menu on a touchscreen display. Knobs and buttons provide physical feedback when they are pressed or turned, which can make it easier to interact with the controls without taking your eyes off the road.
Visibility: In some cases, knobs and buttons can be easier to see and use in bright sunlight or other challenging light conditions, as they do not suffer from glare or reflections in the same way that a touchscreen display might.
Safety: Using physical knobs and buttons can be safer than interacting with a touchscreen display, as it allows the driver to keep their hands on the wheel and their eyes on the road.

Therefore, it is crucial to have a simplified human-machine interface (HMI) on the central screen of a car that is user-friendly, reliable and intuitive in order to minimize the learning curve for drivers and enable them to easily and efficiently access the desired features without encountering any errors. The most important of these is the virtual voice assistant.

There are several popular virtual voice assistants available in the market today, like Amazon Alexa, Google Assistant, Apple Siri, Microsoft Cortana, Samsung Bixby, Baidu Duer and Xiaomi Xiao AI. In addition, there are other proprietary virtual voice assistants designed specifically for the automotive industry, such as Cerence, SoundHound Houndify, Harman Ignite and Nuance Dragon Drive.

The majority of these virtual assistants in the automotive industry are created to seamlessly integrate with the vehicle infotainment systems to offer drivers a variety of voice-activated functionalities, including hands-free phone calls, weather updates, music streaming, and voice-activated navigation. Moreover, they are designed to recognize and respond to natural language commands, enabling drivers to engage with their vehicles in a more intuitive and effortless manner. By providing a safe and convenient way to interact with vehicles, these virtual voice assistants allow drivers to keep their hands on the wheel and eyes on the road.

While virtual voice assistants have improved significantly in recent years, there are still some challenges that need to be addressed. Here are some common problems that currently exist with virtual voice assistants:

Understanding complex commands: Virtual voice assistants may encounter difficulties in comprehending intricate commands or requests that involve several variables or conditions.
Accents and dialects: Virtual voice assistants may also have difficulty understanding users with different accents or dialects.
Background noise: Virtual voice assistants can be sensitive to background noise, which can make it difficult for them to understand user commands or requests.
Privacy concerns: As virtual voice assistants become more ubiquitous, there are growing concerns about the privacy of user data.
Integration with other automotive systems: Virtual voice assistants may have difficulty integrating with other systems or devices, which can limit their functionality and usefulness.

ChatGPT can speak the natural language and converse like a human because it is a language model that has been trained on a massive amount of text data using a deep-learning technique called transformer architecture. During its training, ChatGPT was exposed to vast amounts of natural language text data, such as books, articles and web pages. This allowed it to learn the patterns and structures of human language, including grammar, vocabulary, syntax and context.

Unlike broad-based training methods, natural language training, such as that offered by ChatGPT, allows for the development of models that are finely tuned to specialized data sets, which may include frequently used vehicle commands or a range of distinct national accents. The model is then fine-tuned by further training it on the large corpus of unlabeled data to improve its language understanding capabilities.

The following figure shows our forecast for the use of intelligent voice control in cars.

Source: Global Automotive ADAS/AD Sensor Forecast by the Level of Autonomy, 2021-2030F

Overall, the potential of natural language voice conversation assistants in cars is vast, and with ongoing research and development, we can expect to see more advanced and sophisticated voice assistants in the future. Developing a successful natural language virtual voice assistant for use in cars is a complex and time-consuming process that requires multiple iterations of training and fine-tuning.

Since the development necessitates a considerable amount of data, computational resources and expertise, only a handful of companies such as Microsoft, Tesla, NVIDIA, Qualcomm, Google and Baidu have the resources to undertake this work. The development of the technology is estimated to take three to four years. There will be an increased demand for vehicles above Level 3.

As highlighted in our report “Should Automotive OEMs Get Into Self-driving Chip Production?”, the automotive industry will confront obstacles related to electrification and intelligent technology, necessitating sustained capital investments and support from semiconductor suppliers. Consequently, only a handful of established car manufacturers with considerable economies of scale will be able to finance these initiatives. The growing popularity of natural voice control in cars will only intensify these challenges.

Related Blogs:

Related Reports:

Top

Insights Tag： generative AI

MediaTek Strengthens Premium Push With Gen AI Capabilities

Related Posts

Qualcomm Snapdragon 8 Gen 3 Unveiled: On-Device Generative AI Takes Center Stage

On-device Generative AI: The Key Theme

WATCH: On-Device Generative AI: Text-to-Image in Under a Second

These models are being used for a breakthrough in camera experiences like:

Snapdragon 8 Gen 3: Key Specifications

Enhanced Connectivity: 5G Advanced Ready, Wi-Fi 7 and Dual Bluetooth

OEM Partners and Expected Timeline

Qualcomm S7, S7 Pro Gen 1 Sound Platforms with AI-enhanced Audio and XPAN Technology

Snapdragon Seamless: Enabling Interoperability Between Windows and Android Devices

Guest Post: OpenAI Haymaker?

Related Posts

Podcast #70: Qualcomm Driving On-device Generative AI to Power Intelligent Experiences at the Edge

Click the play button to listen to the podcast

Podcast Chapter Markers

Also available for listening/download on:

Podcast #69: ChatGPT and Generative AI: Differences, Ecosystem, Challenges, Opportunities

Click to listen to the podcast

Podcast Chapter Markers

Also available for listening/download on:

Guest post: AI Business Model on Shaky Ground

Related Posts

AI Voice Assistants to Push Success of Autonomous Driving, Software-defined Vehicle

Coverage

Insights

Social media

Contact us

Top

Related Posts

On-device Generative AI: The Key Theme

WATCH: On-Device Generative AI: Text-to-Image in Under a Second

These models are being used for a breakthrough in camera experiences like:

Snapdragon 8 Gen 3: Key Specifications

Enhanced Connectivity: 5G Advanced Ready, Wi-Fi 7 and Dual Bluetooth

OEM Partners and Expected Timeline

Qualcomm S7, S7 Pro Gen 1 Sound Platforms with AI-enhanced Audio and XPAN Technology

Snapdragon Seamless: Enabling Interoperability Between Windows and Android Devices

Related Posts

Click the play button to listen to the podcast

Podcast Chapter Markers

Also available for listening/download on:

Click to listen to the podcast

Podcast Chapter Markers

Also available for listening/download on:

Related Posts

Coverage

Insights

Social media

Contact us

Term of Use and Privacy Policy