据CNBC报道，TikTok CEO周受资预计，AI将在未来三年内推动重大科学突破，并给众多行业带来变革。周受资周一在新加坡举行的星展银行亚洲前瞻峰会上表示：“未来两到三年，我们将在物理学、生物技术、化学和材料科学领域取得真正的科学突破，而这一切都将得益于AI。这是非常、非常、非常有可能发生的。”
周受资
就在周受资发表上述言论之际，TikTok母公司字节跳动正将AI研究拓展至消费级应用以外的领域。今年4月，字节跳动旗下Seed研究团队发布了Protenix-v2，这是一款专注于生物分子结构预测与设计的AI模型，在药物发现方面具有潜在应用价值。
周受资表示，AI正在快速进步，它未来能做到的事情，将远远不只是现在这些用途。与此同时，如何确保日益强大的系统在人类伦理边界内运行，仍然是一个有待解决的问题。
“如果这种情况发生，那么问题就来了：如何让模型与人类的价值观保持一致？如何让机器按照我们人类认为理所当然的方式行事？”周受资称。
长期影响被低估
他强调，技术发展正在加速，尤其是在科学研究领域。他认为，AI的长期影响可能远超目前的预期。
“趋势非常、非常明显。我们才刚刚开始探索AI可能具备的能力，”他说，“如果你看看趋势走向、进展所在，尤其是在科学领域，它此刻正沿着一条非常陡峭的曲线攀升。”
周受资解释说，虽然AI的短期影响可能被夸大，但其长期潜力仍被低估。他说：“我认为，我们可能高估了AI的短期影响，却低估了它的长期潜力。”
周受资所展现出的乐观态度，正值科技巨头正斥资数十亿美元投入AI驱动的科学研究之际。上周，美国政府、Meta和Google宣布参与一项耗资18亿美元的计划，旨在开发能够预测细胞行为的AI模型，以加速生物学研究和药物发现。
TikTok的母公司字节跳动也在开发自己的AI模型，并探索与科学家合作开展研究的机会。
“我们正在努力探索如何与众多科学家合作，共同推动科学进步，”周受资补充道，“我们在这方面已经做了一些工作，非常令人兴奋。”
他将科学发现描述为AI最有前景的应用领域之一。与此同时，AI也正在日益重塑TikTok的运营方式，涵盖内容创作、内容推荐、搜索和内容审核等多个方面。
更强安全保障
周受资警告说，随着AI能力不断增强，尤其是自主运行的AI系统得到更广泛应用，行业必须采取更强有力的安全防护措施。
“我认为网络安全风险相当大。”周受资称。他警告说，用户可能会在未充分理解潜在后果的情况下，授予AI智能体广泛的权限。
他指出，允许AI智能体自主进行购物存在风险，因为这些系统在急于完成任务时，最终可能购买超出用户本意的物品。
周受资还强调，企业面临的网络安全挑战日益严峻，因为随着AI模型变得越来越先进，恶意行为者实施欺诈活动的成本正大幅下降。
“因此，你需要有合适的工具，来应对制造问题的成本迅速下降所带来的威胁，包括给我们这样的平台造成的安全问题。”他表示。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
According to CNBC, TikTok CEO is projected to drive major scientific breakthroughs and bring change to a wide range of industries over the next three years. At the Starfleet Bank Asia Vision Summit, held in Singapore on Monday, it was stated: “We will achieve real scientific breakthroughs in physics, biotechnology, chemistry and material science in the next two to three years, all of which will benefit AI. This is very, very, very likely.”
Financial support
As TikTok parent company, byte, was making this statement, extending AI research beyond consumer-level applications. In April of this year, the byte jump-starting team of Feed publish took Protenix-v2, an AI model focused on biomolecule structure prediction and design, with potential application value for drug discovery.
Zhou expressed the view that AI was making rapid progress and that what it could do in the future would be far more than just those uses. At the same time, ensuring that increasingly powerful systems operate within the ethical boundaries of humankind remains an outstanding issue.
“If that happens, then the question arises: how can models be aligned with human values? How can machines be allowed to act as we humans take for granted?”
Long-term impacts underestimated
He stressed that technological development was accelerating, particularly in the area of scientific research. In his view, the long-term impact of AI could be far greater than currently anticipated.
“The trend is very, very clear. We are only just beginning to explore the potential capabilities of AI,” he said, “If you look at trends, where progress has been made, especially in the field of science, it is climbing along a very steep curve.”
It was explained that while the short-term impact of AI might be exaggerated, its long-term potential was still underestimated. He said, “I think we may have overestimated the short-term impact of AI but underestimated its long-term potential.”
The optimism that has been demonstrated by weekly funding comes at a time when the technology giant is investing billions of dollars in AI-driven scientific research. Last week, the United States Government, Meta and Google announced their participation in a $1.8 billion programme to develop an AI model that can predict cell behaviour to accelerate biological research and drug discovery.
TikTok ' s parent company, byte, is also developing its own AI model and exploring opportunities for collaboration with scientists for research.
“We are trying to explore ways to work with many scientists to advance science”, supplemented by “We have done some very exciting work in this area”.
He described scientific discoveries as one of the most promising applications of AI. At the same time, AI is increasingly reshaping TikTok ' s operating methods, covering a wide range of content creation, content referral, search and content auditing.
More security.
Zhou warned that with the increasing capacity of AI and, in particular, the wider application of self-functioning AI systems, industry must take stronger safety precautions.
"I think cyber-security risks are quite high." The week has been described. He warned that users might grant AZAI agent extensive permission without fully understanding the potential consequences.
He pointed out that there was a risk that AZAI agent would be allowed to shop autonomously, as those systems might eventually purchase items that were beyond the user ' s intent when they were anxious to complete their tasks.
It was also highlighted that the network security challenges faced by businesses were increasing, as the costs of fraudulent activities by malicious actors were decreasing significantly as the AI model became more advanced.
“You therefore need to have the right tools to deal with the threat posed by the rapidly falling costs of creating problems, including the security implications for platforms like ours.” He said.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.