Chuyển hướng sang AI: cơ hội hay rủi ro cho các startup?

• Kamarul A Muhamed, người sáng lập Aerodyne, đã chuyển hướng công ty từ doanh nghiệp drone sang giải pháp AI dựa trên dữ liệu thu thập từ drone.

• Aerodyne hiện có hơn 600 nhân viên, trong đó 150 người làm việc về khoa học dữ liệu, kỹ sư AI và phát triển phần mềm.

• Công ty đang phát triển mô hình ngôn ngữ lớn (LLM) cấp doanh nghiệp để tích hợp dữ liệu drone với các hệ thống quản lý và tự động tạo báo cáo.

• Ứng dụng đầu tiên sẽ là giám sát và bảo trì các tháp truyền tải điện, dự đoán vấn đề bảo trì và lên lịch sửa chữa trước.

• Theo dữ liệu từ Crunchbase, 37% tổng vốn đầu tư từ 12/5 đến 12/7/2023 (12,9 tỷ USD trong 589 vòng gọi vốn) đã đổ vào lĩnh vực AI.

• Denning Tan, đối tác tại GenAI Fund ở Malaysia, ước tính 95% startup không tập trung vào AI ở Đông Nam Á đang tích hợp chức năng AI vào sản phẩm.

• Việt Nam, Singapore và Malaysia đang dẫn đầu trong việc áp dụng công nghệ AI trong khu vực.

• Các nhà đầu tư hiện kỳ vọng thấy AI trong các sản phẩm mới, dù chỉ là tính năng đơn giản như chatbot.

• Nhiều công ty lớn như The Washington Post và Dell cũng đang chuyển hướng sang AI.

• Dash Dhakshinamoorthy, chuyên gia hệ sinh thái startup, cho rằng việc chuyển hướng sang AI có thể tốt nếu giải quyết được vấn đề thực tế của khách hàng.

• Aerodyne đã tăng doanh thu đáng kể sau khi chuyển hướng sang dịch vụ bảo trì bằng drone và AI.

• "AI washing" - tuyên bố sai về việc sử dụng AI - đã trở thành vấn đề pháp lý ở Mỹ, nhưng chưa phổ biến ở Đông Nam Á.

• Thị trường chứng khoán gần đây đã chứng kiến sự sụt giảm do lo ngại về bong bóng AI, xóa sổ gần 3 nghìn tỷ USD giá trị thị trường.

• Các nhà phân tích dự đoán việc gọi vốn cho các dự án AI sẽ trở nên khó khăn hơn trong thời gian tới.

📌 Xu hướng chuyển hướng sang AI đang lan rộng trong giới startup, với 37% vốn đầu tư đổ vào lĩnh vực này. Tuy nhiên, các công ty cần thận trọng để tránh "AI washing" và tập trung vào giải quyết vấn đề thực tế. Thị trường đang trở nên thực tế hơn, đòi hỏi các startup phải liên tục học hỏi và thích nghi.

https://www.techinasia.com/pivot-pivot-ai-question-startups

To pivot or not to pivot to AI: that’s the question for startups
In 2016, Kamarul A Muhamed read a report that the drone industry would be worth US$128 billion by 2020.

Kamarul was ecstatic. He had founded Aerodyne, a Malaysian drone company based in Cyberjaya, just two years earlier.

“So I was celebrating,” he told the crowd at the Tech in Asia Conference in Kuala Lumpur in July. He studied the paper, reading it back to back several times, and even flew to London to meet the writer.

But Kamarul soon learned that the report’s assertions were “far from the truth,” he tells Tech in Asia in a follow-up interview. He says that demand failed to materialize from the ecommerce sector and government regulations proved to be an obstacle.

“Fast forward to 2020, I don’t think it’s even 25% of the way [to US$128 billion],” the founder adds.
But in 2017, Kamarul noticed something else. Drones were not only capable of offering a service, such as security or delivery, but were also a prime source of data. That sent him working on a data solution that same year.

Seven years later, Kamarul says that Aerodyne has successfully pivoted, focusing on AI solutions based on the data that the drones provide. He describes the firm as “more of a data science company that uses data from drones to provide predictive analysis and enterprise resource planning solutions.”
The company now has over 600 people with a team of “150 data scientists, AI engineers, and software developers” working on data.

Kamarul says the firm will now be focused on combining data from the drones with an enterprise-level large language model (LLM) that will integrate data with procurement and finance as well as automatically generate reports.

The first application will be with electrical transmission towers. The drones will monitor and photograph the towers, and the LLM will include industry-specific information relevant to maintenance and operations of the towers.
He adds that this AI-driven system would be able to forecast maintenance issues and allow the firms to schedule repairs and order parts in advance.

But is AI really needed to do this? Can’t maintenance be done using existing software?
Kamarul admits that AI isn’t the only way these maintenance issues can be addressed, but he says the LLM removes the need for a human, who would have to act as an intermediary between the various systems.

“[People] need to dig out the information, [they] need to build the intelligence, and [they] need to write the report itself,” he points out. The LLM, in contrast, “understands what you’re asking for.”

AI largest recipient of funding
Aerodyne’s pivot isn’t unusual. Over the last year, more and more companies are looking at moving toward AI as investment in the sector has skyrocketed. According to data from Crunchbase, 37% of all funding raised from May 12 to July 12 of this year – US$12.9 billion in 589 funding rounds – went to AI, making it the “largest single category of business funded by VC dollars.”

For Aerodyne, it has had four funding rounds, the most recent being a series B raise in September 2022.

Denning Tan, a partner at GenAI Fund in Malaysia, estimates that almost 95% of non-AI centric startups in Southeast Asia are building AI functionalities into their products today. He tells Tech in Asia that Vietnam and Singapore, and to some extent Malaysia, are leading the way in embracing the tech.

Tan adds that investors are expecting to see AI in new products today, even if it is just a very simple addition such as a chatbot.

“If you don’t have it, then maybe it raises a question mark,” he says.

According to Tan, most startups have the capabilities to very quickly develop wrappers, a small application that allows a user to access someone else’s AI via an API, as part of their offerings. But the ability to integrate AI into core products and then use that to attract investment is still a ways away.

“The ones who are truly building something new on a very vertical basis – that is still a work in progress,” he points out.

The pivot to AI is also happening in established firms, from The Washington Post to Dell. Even bitcoin miners are jumping on the trend.

While Aerodyne’s pivot to AI appears to be successful, is such a move a good idea for startups? Dash Dhakshinamoorthy, a startup ecosystem veteran with over 25 years of experience helping new firms in Malaysia, thinks it can be – if the company is pivoting to a product or service that users are willing to pay for to solve a problem.

“But don’t chase it just because it’s technology,” he tells Tech in Asia.

“There are too many products that are chasing after customers,” he explains. “They’re just saying, ‘Oh, this is sexy, I’m going to build it and then let me see whether somebody will buy it’.”

From days to minutes
When Aerodyne pivoted to AI, Kamarul already had a problem he wanted to fix: making maintenance of hard assets, such as telecom towers and pipelines, more efficient by using drones.

He started off with power lines in Indonesia. Normally, power companies need to send out teams by helicopter to inspect and repair the lines and towers, inspecting them for problems such as structural damage, overgrown vegetation, and unplugged cables.

This could take a human team three to five days, from start to finish, for a single tower due to travel time – towers are often in remote locations – and the time it takes to source parts.

A drone can do the same inspection in just 50 minutes by taking pictures of the tower and, through an algorithm, translating this into an accurate digital twin of the structure. The twin is then used by the repair crew to determine when future maintenance will need to be done.

But with the launch of the LLM, Kamarul says clients will be able to query the conditions of their assets easily, rather than only finding out on inspection.

Even before the LLM is introduced, the founder says revenue from drone maintenance services is on the rise. Aerodyne acquired Australian-based Sensorem in 2020, when it had revenue of A$125,000 (US$83,000). In 2023, revenue jumped to A$4 million (US$2.6 million).

That same year, Aerodyne also acquired FEDS (Falcon Eye Drones). Revenue in 2020 was 3 million dirham (US$817,000), which grew to 10 million dirham (US$2.7 million) last year.

Kamarul says the initial investment in LLM development stands at under US$10 million, and the first version of the model is projected to be completed within six months. A public release is expected in 12 to 18 months.

AI washing
But companies also need to be careful about delivering on what they promise. Over the last few months, one unfortunate trend – AI washing – has become a legal issue. This is a term used for companies who falsely claim to have added AI to their products.

In March, two US investment firms agreed to pay penalties in a case brought by the US Securities and Exchange Commission.

According to the SEC’s statement
Toronto-based Delphia promised its AI could “predict which companies and trends are about to make it big and invest in them before everyone else.” The SEC said the firm did not in fact have such technology.

The SEC also accused San Francisco-based Global Predictions of making false and misleading claims in 2023 that it has the “first regulated AI financial advisor” and could provide “expert AI-driven forecasts.”

Neither firm admitted or denied the SEC’s findings. Delphia will pay a fine of US$225,000 while Global Predictions will pay US$175,000.

SEC Chair Gary Gensler said the agency had “seen time and again that when new technologies come along, they can create buzz from investors as well as false claims by those purporting to use those new technologies.”

GenAi Fund’s Tan says that AI washing hasn’t yet become a major issue in Southeast Asia.

“I have seen cases, but it went away pretty quickly because the investors were just too smart about it,” he shares, adding that investors in the region generally have had difficulty accurately pricing what the premium is for AI and what the associated risks are. Because of that, they move slower when evaluating new investments in the sector. They have also learned from watching the mistakes made by other AI companies in the US.

Tan also says that investors in Southeast Asia watch the speed at which a startup pivots to AI. A quick pivot can raise questions.

Lingering concerns
But the rise of AI washing, combined with a market perception that capital expenditures are outpacing any immediate hopes of revenues, has prompted a backlash from some investors.

In July, Wedbush analyst Dan Ives wrote that “heading into earnings season in mid-July, the Street needed to see validation data points from the tech world supporting this AI revolution thesis … and importantly show monetization and use cases are on the doorstep.”

That didn’t happen, and on August 5, the public markets experienced a huge fall, driven in part by concern that AI was in a bubble. Almost US$3 trillion in market value was erased from the industry in under a month, although it has since recovered somewhat.

Despite the recovery, market analysts expect that funding for projects, especially those with AI, will become tougher.

Dhakshinamoorthy says the market is just “getting real” and that founders need to get back to fundamentals.

“We are on the journey to irrelevance all the time as companies,” he points out. “That means pivoting is an inevitable thing. If AI is what you’re going to do … you need to be a learner. A startup entrepreneur has to be a learning machine.”

Currency converted from Australian dollar and United Arab Emirates dirham to US dollar: US$1 = A$1.51; US$1 = 3.67 dirham.