As one of the world's most influential science and technology think tanks, MIT Technology Review, with its unique perspective and professional insight, has always been at the forefront of technological innovation. Since 2001, it has started the selection of "10 Breakthrough Technologies" to discover not only the superficial technological progress but also the technological breakthroughs with transformative potential.

With new discoveries frequently hitting the headlines and major technology companies constantly introducing new products, it's extremely difficult to determine which breakthroughs can truly stand the test of time. Precisely for this reason, this list is carefully compiled every year. Although the future can't be predicted, it's expected that these selected technologies will continue to shape and influence the world in the coming decades.

Here, we will introduce these ten breakthrough technologies to you in detail in two articles. The following are the first five technologies.

1.Small Language Models

Significance: Large language models have unleashed the powerful potential of AI, and now it's the era of efficient application of small models.

Major Participants:Allen Institute for Artificial Intelligence, Anthropic, Google, Meta, Microsoft, OpenAI

Maturity Stage: Now

It's undeniable that in the field of artificial intelligence, the scale of models is of crucial importance. GPT-3 launched by OpenAI in 2020 was the largest language model at that time. Its appearance directly promoted a leap in AI performance and triggered a technological wave guided by larger scale. As Noam Brown, a research scientist at OpenAI, said at the TEDAI San Francisco Conference (The AI Conference About The Impact and Power of AI) in October 2024, "The huge progress in the field of artificial intelligence in the past five years can be summed up in one word: scale."

However, as the marginal benefits of high-end models gradually decrease, researchers have begun to explore how to achieve "major breakthroughs" through "small scale". In some tasks, small models that focus on specific datasets can perform on a par with or even outperform larger models. This is undoubtedly good news for enterprises that hope to deploy AI in limited fields. If your requirement is for highly repetitive specific requests, you don't need the data of the entire Internet to train the model. Nowadays, most of the tech giants have launched small versions of their flagship models. For example, OpenAI offers GPT-4o and GPT-4o mini. Google DeepMind has released Gemini Ultra and Gemini Nano. Anthropic's Claude3 is divided into the large Opus, the medium Sonnet, and the small Haiku. In addition, Microsoft is also actively developing a series of small language models named Phi.

Moreover, more and more small enterprises have also started to provide small models. The AI startup Writer claims that the performance of its latest language model is on a par with that of top large models in many key indicators, while the number of its parameters is only 1/20 of that of the latter. This efficiency advantage enables small models to be trained and run faster and also provides enterprises with a more economical option. Meanwhile, it's also a great boon for the environment: the computing resources required by small models are significantly reduced, so they consume less energy.

In addition, small models have good portability - they can run directly on our devices without relying on cloud requests. Miniaturization is becoming a new trend in the development of AI.

2. Vera C. Rubin Observatory

Significance: This giant telescope is equipped with the largest astronomical digital camera in history, aiming to quickly and efficiently take a large number of photos and preserve them for a long time.

Major Participants: SLAC National Accelerator Laboratory of the U.S. Department of Energy, the U.S. National Science Foundation

Maturity Stage: In 6 months

When you look up at the night sky, you might as well think about the following fact: The matter we can see only accounts for about 5% of the universe. Astronomers believe that dark energy and dark matter make up the remaining 95%, but what exactly are these mysterious substances? To unravel this mystery, as well as other unsolved mysteries of the universe, a giant telescope named after the American astronomer Vera Rubin will conduct explorations in Chile. Vera Rubin observed in the 1970s and 1980s that in the outer regions of dozens of spiral galaxies, the movement speeds of stars were faster than expected. Her calculations provided important evidence for the existence of dark matter - a substance that we cannot directly observe but obviously affects the orbits of stars and the structure of the universe.

The Vera C. Rubin Observatory will continue this research with higher precision. It is jointly operated by the SLAC National Accelerator Laboratory and the U.S. National Science Foundation and will be equipped with the largest digital camera in the field of astronomy so far. Its first task is to complete the "Legacy Survey of Space and Time". Astronomers will point this giant lens at the starry sky in the southern hemisphere and repeatedly photograph the same area of the sky for ten consecutive years.

By the end of the task, this 320-million-pixel camera is expected to record 20 billion galaxies and collect as much as 60 petabytes of data - three times the storage capacity of the current Library of Congress in the United States. Through the use of specialized algorithms and supercomputers, these images will be integrated into a time-lapse astronomical view. So much information about the distribution and shapes of galaxies will help scientists study the gravitational effects of dark matter and create the most detailed three-dimensional map of the Milky Way to date.

If everything goes smoothly, this telescope will capture its first science-grade image in the middle of 2025, and this important moment is called "first light". Soon after that, the public is expected to see the first photo released by the Vera C. Rubin Observatory.

3. Long-acting HIV Preventive Drugs

Significance: A brand-new drug has shown excellent efficacy in preventing HIV infection and only requires injection twice a year.

Major Participants: Gilead Sciences, GSK, ViiV Healthcare

Maturity Stage: 1 to 3 years

In June 2024, the results of a new drug trial for preventing HIV were announced, which were astonishing. Lenacapavir is a drug that needs to be injected once every six months. In trials in Uganda and South Africa, it successfully protected more than 5,000 women and girls from HIV infection, with an effectiveness rate reaching 100%.

This drug developed by Gilead has multiple advantages. Since 2012, we have had effective pre-exposure prophylaxis (PrEP) drugs for HIV, but these drugs need to be taken daily or before possible exposure to the virus. This is a heavy burden for healthy people. In addition, since these drugs can also treat infections, taking them can bring a certain social stigma. For some people, these drugs are expensive or difficult to obtain. In the clinical trials of Lenacapavir, researchers found that injecting this new drug had a better effect than taking PrEP drugs daily, which was most likely because the participants failed to take the medicine every day.

In 2021, the U.S. Food and Drug Administration (FDA) approved another long-acting injectable drug for preventing HIV called Cabotegravir. This drug is produced by ViiV Healthcare (which is mainly controlled by GSK) and needs to be injected once every two months. However, despite the huge demand, its promotion has been progressing very slowly.

Scientists and social activists hope that the situation of Lenacapavir will be different. Currently, the U.S. Food and Drug Administration has only approved its use for treating HIV patients who are resistant to other drugs. However, Gilead has signed licensing agreements with multiple manufacturers and plans to produce generic drugs for preventing HIV in 120 low-income countries.

In October 2024, Gilead announced more trial results of Lenacapavir. In the trials for preventing HIV infection with the participation of more than 3,200 participants of different genders and sexual orientations (including cisgender gay men, bisexual men, transgender men and women, and other non-binary individuals), its effectiveness rate reached 96%.

The United Nations has set an ambitious goal: to end AIDS by 2030. However, there are still more than 1 million new HIV infection cases worldwide every year. Nevertheless, we already have the drugs needed to achieve this goal. What is most needed now is to make these drugs truly accessible to all those in need.

4. Generative AI Search

Significance: Generative AI is redefining internet search and can also enhance the efficiency of our information retrieval on mobile phones.

Major Participants: Apple, Google, Meta, Microsoft, OpenAI, Perplexity

Maturity Stage: Now

Google's AI Overviews feature launched through its Gemini language model is changing the internet search experience for billions of users. This generative search technology might be the first step towards an AI intelligent assistant - a future tool that can answer all questions and complete various tasks.

Unlike traditional search results which provide a list of links, AI Overviews directly offer users concise answers, helping them quickly gain insights without having to repeatedly click on multiple sources. After its initial launch in the United States in May 2024, due to some inaccurate and even absurd answers that sparked controversy, Google imposed restrictions on the sources of the generated content and stopped using information from user-generated content, satirical or humorous websites.

However, the rise of generative search is not limited to Google. In 2024, Microsoft and OpenAI also launched similar features one after another. Meanwhile, its application scenarios have been continuously expanding. Now, AI-assisted search can analyze images, audio, and video on computers and various devices and provide customized answers.

Despite the increasing number of competitors, Google, by virtue of its dominant position in the global search field, has already promoted the AI Overviews feature to more than 1 billion users. This makes the search experience more like a conversation. Both Google and OpenAI have reported that generative search has changed the way users interact with search engines - people will ask longer questions and make more follow-up inquiries.

The application of this AI technology has a profound impact on the online advertising and media industries. Generative search often summarizes the information of online news or articles in its answers, which may reduce users' willingness to click on the original sources and undermine the advertising revenue of these websites. Some publishers and artists have already sued relevant companies regarding the content training of AI models, and generative search may become a new focus of contention between the media and tech giants.

5.Cattle Belch Inhibitors

Significance: A cattle feed additive that can significantly reduce agricultural emissions has finally emerged, providing a solution to the problem of climate change.

Major Participants:Blue Ocean Barns,DSM-Firmenich, Rumin8, Symbrosia

Maturity Stage: Now

Cows burping is one of the most intractable problems contributing to climate change, and now a substantial breakthrough has finally been made in its solution.

During the digestion process, cows belch out methane, and the same is true for sheep and goats. This potent greenhouse gas is the largest source of emissions from the livestock industry.

According to different analyses, emissions from the livestock industry account for 11% to 20% of the total global climate pollution.

With the growth of the global population and the increase in wealth, the demand for foods such as hamburgers, steaks, butter, and dairy products will continue to grow. It's very difficult to substantially reduce emissions simply by reducing consumer demand.

As a result, cattle belch inhibitors have come into being. DSM-Firmenich, headquartered in the Netherlands, has developed a feed additive called Bovaer, which is claimed to be able to reduce methane emissions from dairy cows by 30% and has an even better emission reduction effect on beef cattle. Its principle is to inhibit an enzyme in the cow's intestine that usually converts hydrogen and carbon dioxide produced during the digestion process into methane that is expelled when the cow belches. In May 2024, the U.S. Food and Drug Administration approved the use of this additive in the United States. Currently, Bovaer has been launched on the market in 55 countries, including Australia, Brazil, and member states of the European Union.

Meanwhile, start-up companies such as Blue Ocean Barns, Rumin8, and Symbrosia are developing additives made from red seaweed, and these products may further reduce methane emissions. Some other institutions are seeking more long-term solutions by attempting to develop vaccines or change the microbial community in the cow's intestine.

However, it's still unclear how many farmers are willing to pay for these products. According to Elanco, the U.S. marketer of Bovaer, farmers who use the product can obtain greenhouse gas credits, and some companies will buy these credits in the voluntary carbon market to reduce the carbon footprint of their enterprises. Meanwhile, Rumin8 says that cattle herds using its additive can provide more meat and dairy products.

Of course, these additives cannot solve all the problems. The livestock industry still needs to take other major measures, such as stopping the encroachment on carbon-absorbing forests, to reduce its climate emissions. In addition, to truly reduce demand, food companies need to develop better, cheaper, and more environmentally friendly alternative products, including plant-based hamburgers and dairy products.

Nevertheless, cattle belch inhibitors have been regarded as an important step in solving this major climate problem.

To be continue,if you want to know the other five breakthrough techologies,click here!