Abigail Hing Wen is the data-centric tech evangelist and AI podcast host for Intel. Beside her work at Intel, she has multiple identities: a lawyer in venture capital, a leader in emerging AI tech, a New York Times bestseller novelist, a wife and a mother of two. One of her novels, "Loveboat, Taipei," drew from her experience participating in a youth camp in Taiwan more than 20 years ago, and is now being adapted into a Hollywood film.During her recent stay in Taipei as a guest speaker at Anchor Taiwan's "Women in Venture Roundtable," Digitimes had the opportunity to talk to her about her career changes, the prospects of AI, and the potential of Taiwan to become a hotbed for startups.Q: You have been very successful in your career path as a lawyer, but what made you decide to transition into a venture capitalist?A: I started off practicing law in Washington, DC, working for the government. Washington DC is a one-industry town, the practice of law. But my husband was working in technology, he would come to the party, talking about all the latest technology, and everyone thought he was so nerdy. He had always wanted to work in the Silicon Valley, but as soon as we got married, he went with me to New York so I could finish law school, and then to DC, so I could clerk in the DC Circuit. So it was kind of my turn (to sacrifice), and it was a hard move for me. DC was my town as a lawyer, but Silicon Valley is also a one-industry town, his town. My husband is working for Google now. After giving birth to my second child, I stayed home taking care of my kids for three years, and I wrote two novels during that time. After that, I decided I need to go back to law, to continue my training. Because Silicon Valley's bread and butter is technology, I went into venture-capital law. The opportunity came very quickly, I went to the company and loved it. Very soon I out-grew my legal job, the work itself is very easy for me, I closed so many deals very fast. There was one year we counted the number of cases and mine was twice as many as my colleagues'. So I thought, I really need to do something else, because I felt I needed to be challenged more. I started moving more towards the business side of things. Three years ago, I connected with an artificial intelligence (AI) company that Intel acquired for US$400 million. They were building a business unit at Intel, for artificial intelligence. And I joined the office of the CTO in a business role focused on incubating new technologies within a company, versus a venture capitalist investing in tech outside of a company.Q: So how do you structure your portfolio? Do you decide specifically to focus on machine learning, or natural language processing, etc?A: Intel invests very broadly. Its mandate has shifted over the years from strategic to financial investment in startups. As a strategic investor, you want to go ahead of the curve to invest in a company, and think about, where should the company be technology-wise in the next 5-10 years. I specifically asked to do AI work, because I have been interested in that and actually have been writing about it in my novels. I really think it's one of the technologies that changes our future. And I want to be part of that.In terms of what we invest in, it depends on whether we are a strategic or financial investor and what's the best for the company in that space. I worked with Andrew Ng's company, Landing.ai, which is in manufacturing, and we have another one called Datarobot. And there are some startups on the human-resources side in the portfolio. We have another startup venture which was just acquired, focused on denoising speech.You meet so many entrepreneurs, and we have a lot of investors who are constantly cultivating relationships with entrepreneurs. They bring in whatever looks interesting from a financial investor's point of view. AI, for Intel, has some advantage as accelerator. Intel is so big, and so diverse, so it is often driven by interests within the company. In my CTO team, one area we focused on was privacy-preserve machine-learning technologies. Because the future in AI is data, we want to make sure data is processed in a way that protects your privacy, and make technologies that comply with regulations. It is a nexus that combines my legal background and my work with technology.Q: You mentioned you need to look 5-10 years ahead as a strategist. What are the AI technologies that are going to change the way we work, the way we live?A: In my podcast, I interviewed Andrew Ng, who said he tried to find one sector that would not be affected by AI, but he could not. They thought maybe haircuts would not be affected, but actually even haircuts would be disrupted. All of it, everything is going to change. We are already using AI in our cellphones, our home devices. They collect big data in order to analyze trends, across all sectors, including finances, healthcare, etc. AI will accelerate our research in so many different fields. I personally am very interested in the intersections of these technologies, like natural language processing with computer vision, recognition, recommendations. I can actually pull them together on the entertainment side. I am thinking about film tech. But that's more likely to be a 5-year plan, still very early.I have a friend who is doing virtual models. I am fascinated to see how she moves from still images to video. Over time I am excited to see when we can create virtual videos using AI technologies. Right now we do animation, which is like analog in comparison. But we are moving toward virtual humans. Digital Domain and Facebook are just a few companies working on it. That's the holy grail of that space. But we are still at very early stages.I was a beta tester of OpenAI's GPT-3 language model, which is one of the largest language models in existence now. Actually I played off it a bit when I was writing my second book. What I did was putting into a few sentences that I was thinking about writing. It was able to generate text by predicting the next word and the sequences, given all the words that came before it. It is pretty good: it's cogent, the grammar is good, tone actually matches the tone I put into it. If I put in "wealthy family, succession story", it kind of develops a generic story around that. It is useful to me, because it shows me this is how the generic writer might write, so I am not going to write that. But that also reminds me what else I need to include in my own book. There are a number of startups that are springing up around GPT-3, and I think that's going to be very exciting to see what those applications are.Q: But as a writer, are you worried that the AI robots will replace you?A: No, not at all. When I interviewed Jerome Pesenti, the VP of Facebook AI Research, we discussed this topic about AI and creativity. AI is just a tool to enable us to make better art. It's like a camera, which doesn't replace watercolors. Those are two different mediums, but there might be more photographers than watercolorists now. And I think AI is going to be a tool for all art creators. And AI art is being generated out there. Art is like a two-way street. There is the creator and there is the consumer. And the consumer completes the artistic experience. They bring their own experiences to bear. So when people look at a piece of AI-generated art and see meaning in it is because of the experiences they are bringing. The human creativity is what distinguishes us from artificial intelligence, and that is going to be difficult to ever be replaced.Q: You are a podcast host and also an AI evangelist for Intel. We all know that Intel is a B2B semiconductor company. Why does Intel feel it's so important to have an evangelist writing articles and using podcast as a medium to communicate to the public about AI?A: This is me speaking on my own, not representing the company. The vision of the podcast is to help build our thought leadership. We have so many brilliant, incredible scientists and researchers of artificial intelligence at Intel. But they are not very visible. People don't naturally think of Intel as an AI company, because it focuses on the underlying chips. But on AI-related issues like ethics, we do have things to say. I love that platform of Intel, because it's agnostic, it doesn't have an agenda to drive AI, the way that some companies might. That gives me a little bit of flexibility to explore a number of topics. The podcast's emphasis is on a C-suite audience and what I call an "NPR" (National Public Radio) audience. They are smart, informed people who may not know too much about artificial intelligence. But I want to make it accessible to them.And part of the vision is that, I do think AI is going to transform the world. I want as many people as possible to know what is going on. That transparency is actually helpful to build trust in technology.Q: How does Intel create synergy with AI in its products?A: There are many cool technologies, but a lot of them are still in the labs. If we look at the investments Intel has in AI technologies, that would be a good indicator of what they are possibly exploring. And there are a lot of companies that are using Intel chips as the base of their AI, or deep-learning applications. But there are always interesting projects that are incubated around AI. And with the maturity of the technology, there will be more demand for chips and computing power. As Andrew Ng puts it, the "appetite for computing power from the AI revolution is infinite."Q: Earlier you mentioned that the number of companies in Intel's startup investment portfolio has decreased from more than 1,200 to 400+, what was the lesson learned to make that dramatic change?A: No, no, it's more like exits, rather than lesson learned. We had lots of exits over the years - companies get acquired, sometimes they closed. In the startup ecosystem, as a general rule, most startups will fail; but as an investor, you want to make the best bets you can and help the startup company to be successful. The portfolio has gone up and down over the years, usually due to IPO, M&A, or the company folding.Q: But why the turn to financial investment from strategic?A: That's a corporate decision. People recognize that corporate venture capital (CVC) can bring to bear the weight of their expertise, but it does sometimes limit you to companies that have some kind of ties to the larger parent company, like Intel. And the very ambitious investors all want the best deals. You have more options as a financial investor, rather than just strategic.Q: As an investor, do you see Taiwan as a potential hotbed for AI startups in the next 5-10 years?A: I am already very impressed with the tech sphere here. With COVID-19, there are so many COVID refugees coming all over the world to Taiwan. That's an incredible opportunity to create a branch hub for innovation. The universities here are fantastic. Taiwan has almost all the ingredients except for one key thing, which is the uptake for risk, being willing to take big bets, smart bets, but make that seem ok to failure. As I said, not all of Intel's huge portfolio all wins. But that is part of the way you get to your wins.(Editor's note: The Women in Venture series is a collaboration with Digitimes' strategic partner Anchor Taiwan, a platform to connect the world with Taiwan focusing on corporate innovation and cross-border expansion. The Women in Venture Roundtable is a network of 80+ female investors and bi-monthly sessions featuring world-class guest speakers. More info: Anchor Taiwan. For the podcast, see Spotify, Apple Podcast or Soundcloud: https://open.spotify.com/episode/0Dy6TKKLIazz8oaCW0oiiE.)Abigail Hing Wen, data-centric tech evangelist and AI podcast host for IntelPhoto: Micahel Lee, Digitimes, September 2020
The freshly released iPhone 12 devices all feature OLED screens. China-based panel maker BOE reportedly was initially not among the OLED suppliers for the new Apple smartphones, but is now said to be ready to ship OLED screens in small volume for the 6.1-inch iPhone 12. The overall handset market in China may have been weak this year, but 5G smartphone sales are picking up. Sales of 5G phones in China may reach 150 million units in 2020. The TV market in China has also been unimpressive. TV sales during the shopping season around China's October 1 National Day dropped on year.BOE to ship OLED panels for iPhone 12 in small volume, say sources: BOE Technology reportedly will start shipping 6.1-inch OLED panels, albeit in small volume, for the just released iPhone 12 in the second half of October, with prospects to gradually ramp up related shipments in 2021, according to industry sources.China 5G phone shipments to top 150 million units in 2020: Sales of 5G phones in China, already in full swing, are expected to gain a further boost in fourth-quarter 2020 thanks to higher than expected pre-orders for iPhone 12 lineup received by e-commerce operators there, according to market observers.TV sales down in China during National Day shopping: TV shipments during China's three-week shopping season around the National Day (September 21-October 11, 2020) totaled 2.684 million units, down 8.3% on year, according to All View Cloud (AVC).
The iPhone 12series are expected to see significant sales, with Apple likely to increase orders for its ecosystem partners. Apple will reportedly increase SLP orders with its PCB suppliers for the iPhone 12 mini, as the smaller-size model has good sales prospects in Asia. In the semiconductor backend service sector, strong demand for wire bonding capacity is likely to last beyond first-quarter 2020. And Foxconn is gearing up efforts for the electric vehicle market. The EMS giant has unveiled its MIH software/hardware open platform for EV development.iPhone 12 mini mainboard suppliers may land extra orders: Taiwan's PCB makers Compeq Manufacturing and Unimicron, both now supplying SLP mainboards for iPhone 12 mini and iPhone 12 Pro Max, are likely to land additional orders from Apple as the former model stands a good chance of seeing better-than-expected sales, according to industry sources.Taiwan IC backend service firms ramping wire bonding capacity: The current tight wire bonding (WB) packaging capacity at Taiwan's IC backend service firms is likely to last beyond the first quarter of 2020 thanks to strong demand from analog and logic IC vendors, according to industry sources.Foxconn unveils software/hardware open EV development platform: A joint venture newly established by Foxconn Electronics and Yulon Motor has unveiled MIH, a software/hardware open platform for developing electric vehicles (EVs), according to Foxconn.
NADI System, Memorence AI and Verdigris Technologies have developed AI-based solutions for manufacturing.NADI has developed a 3D visualization management platform based on PSIM (physical security information management) architecture. The platform integrates data collected from manufacturing systems and IoT devices as 3D visualized displays for users to undertake analysis and prediction, according to company general manager Steven Chang.The 3D displays use in-house-developed algorithms and locational parameters for each production line, Chang said, adding that the platform is intended to help manufacturers attain real-time, accurate and remote management.Memorence has developed Memorence Suite, a GUI (graphical user interface)-based visual deep learning system capable of data management, AI model learning and inference, and results analyses, company founder and CEO Hsiao Pai-heng noted.To adopt AI-based solutions to upgrade production, the first step is to define the problems to be solved and choose appropriate image-capturing devices, and then keep collecting images and data and establish corresponding databases. The final step is to build AI models and let the models keep learning for optimization, Hsiao indicated. Memorence has helped clients hike defect detection accuracy to over 99.9%, Hsiao said.Verdigris, a company established in Silicon Valley by Taiwanese entrepreneurs, has developed AI-based sensors and algorithms to help factories diagnose power use for equipment and manage details of power consumption to maximize efficiency in power use, said company senior sales manager Jacky Wu.Verdigris has helped US-based Katerra install six smart monitoring systems consisting of 216 smart power meters at six switchboards at its factory to monitor 72 different electrical circuits, Wu noted. Verdigris also provides smart early warning and power use prediction during hours of peak power demand.Memorence founder Hsiao Pai-heng (left to right), NADI general manager Steven Chang, and Verdigris senior sales manager Jacky WuPhoto: Digitimes, October 2020
TSMC expects revenues from its 5nm manufacturing process to pick up in 2021 with the share to rise from only 8% in 2020 to more than 20%. Meanwhile, as the IT industry continues witnessing insufficient wafer foundry support, backend houses have also extend their delivery lead times to reflect their tight capacities. Demand from China will strongly influence iPhone 12's sales performance in 2020.TSMC expects 5nm chip sales to boost in 2021: TSMC CEO CC Wei has reiterated his previous remark that the foundry expects to see 5nm process technology to account for 8% of its total wafer revenues this year. The proportion will climb further to nearly 20% or above the level, said Wei at an earnings conference call on October 15.Taiwan backend houses extend delivery lead times: Delivery lead times at Taiwan-based packaging and testing houses have been extended, as fabless chipmakers are striving for capacity support, according to industry sources.China sales crucial for iPhone 12 performance: Sales of the just-released iPhone 12 lineup in China would be a linchpin for Apple to see whether its iconic product line could sustain growth in 2020 and beyond, according to industry observers.
The 5G subscriber market has continued growing steadily in China, with its 5G connected devices in use on track to grow to 185 million units by year-end 2020, having reached 66 million units in the first half, Digitimes Research estimates.After building over 400,000 5G base stations in first half-2020, Chinese telecom operators as a whole will install another 250,000-300,000 such base stations in the second half of the year, Digitimes Research indicates.To mitigate the financial pressure for 5G infrastructure buildups, China Mobile has reached a deal to share its 5G networks with China Broadcasting Network (CBN), while China Telecom has also teamed up with China Unicom for network sharing.Combined capital expenditure of China's top-3 telecom operators – China Mobile, China Telecom and China Unicom – totaled CNY88 billion (US$13.094 billion) in first-half 2020 and is to expand to CNY180 billion by year-end, representing an increase of 338% from a year earlier.Despite heavy investments for 5G network construction, the three telecom operators have yet to see meaningful improvements of their financial results in terms of ARPU (average revenue per user), EBITDA and EBITDA margin for migrating into 5G from 4G.
Apple's new 5G smartphone series is expected to see significant sales momentum, with shipments for iPhone 12 devices standing a chance of hitting as many as 80 million units by the end of 2020, although the US vendor is looking to build an inventory of 71-75 million units for this year. Foxconn is the major assembler for the iPhone 12 series. The Taiwanese ecosystem remains a major partner for Apple, as its PCB makers will supply 60% of the SLPs for th enew iPhones. Shipments of iPhone 12 lineup may top 80 million units by year-end 2020: Shipments of Apple's just-unveiled iPhone 12 lineup may reach as high as 80 million units by year-end 2020, as the vendor's more affordable pricing strategy is likely to entice users with older iPhones to upgrade, according to market observers.Foxconn remains major iPhone assembler: Foxconn Electronics (Hon Hai) has secured the bulk of orders for the new iPhone 12 series and will see its shipments for the new Apple smartphones peak through the first quarter of 2021, according to market sources.Taiwan PCB makers obtain 60% of SLP orders for new iPhones: Zhen Ding Technology, Compeq Manufacturing and Unimicron Technology have together landed nearly 60% of total substrate-like PCB (SLP) orders placed by Apple for its new iPhone 12 series, according to industry sources.
Taiwan has to concentrate resources in developing a smart medical care sector with its own characteristics and international competitiveness, with electronic medical devices and telemedicine among its focus, according to Steve Kuo, president for National Yang-Ming University.For electronic medical devices, resources need to be focused on domain knowledge in selected medical areas and integration with computer-aided interpretation systems and other medical hardware/software, Kuo said.Taiwan-based ITC makers stand a good chance of stepping into electronic medical devices and they can help developers of such devices, Kuo said.Cross-sector integration of domain knowledge concerning medical information, public health, information engineering and other related areas is crucial to development of electronic medical hardware/software, Kuo indicated.Yang-Ming and National Chiao Tung University will be merged in 2021, and the combination of the former's medical expertise and the latter's focus on ITC science and technology is expected to create synergy, Kuo said.National Yang-Ming University president Steve KuoPhoto: Michael Lee, Digitimes, October 2020
Taiwan Association of Machinery Industry (TAMI) and Taiwan Electrical and Electronic Manufacturers' Association (TEEMA) have jointly established a smart manufacturing alliance to help local makers and promote digital transformation, in a bid to turn the country into a high-end manufacturing hub in Asia.The US-China trade war and coronavirus pandemic are changing the global supply chains and digital transformation is inevitable for Taiwan's manufacturing sectors, TAMI said. But about 97% of Taiwan's manufacturers are small- to medium-size ones, and digital transformation means huge costs for them, TAMI noted.The Taiwan Smart Manufacturing Alliance is committed to facilitating digital transformation of equipment makers and small- to medium-size manufacturers, it said, adding the new alliance is so far the largest cross-industry partnership for promoting smart machinery.Inaugural ceremony for Taiwan Smart Manufacturing AlliancePhoto: TAMI
Many IC components have been in short supply due to insufficient foundry support. Now foundry houses are asking their fabless IC clients to book well in advance for capacity for 2021 to make sure they have enough support. The components shortages are already sending driver IC and MOSFET prices rising. Their prices are expected to increase by more than 10% in fourth-quarter 2020. Memory sport market prices are also rising, but SSD and DRAM module sales in China are actually disappointing.Taiwan IC design houses urged to book foundry capacity for 2021: Taiwan-based IC design houses are being advised by their foundry partners to place orders for 2021 in advance to ensure sufficient capacity, according to industry sources.LCD driver IC, MOSFET prices to rise over 10% in 4Q20: LCD driver IC and MOSFET chip prices are set to rise over 10% in fourth-quarter 2020 and will continue their rally in the following two quarters to reflect tight supply and rising manufacturing costs, according to industry sources.Memory module houses see demand in China disappoint: Despite a continued rally in memory spot market prices, channel distributors in China continue to see disappointing SSD and DRAM module sales, according to sources at memory module makers.