亚马逊亮眼业绩背后:跳出最强模型内卷,走出差异化AI突围路径
财联社7月31日讯(编辑 李莹)亚马逊最新发布的财报印证了一个在硅谷日益获得认同的观点:想要在AI赛道取胜,未必非要拥有性能最强的模型。
亚马逊交出了亮眼的业绩,其中最受关注的AWS云服务营收同比大增37%。这一爆发式增长,主要归功于其云平台运行着来自Anthropic、OpenAI等行业领头羊的多种AI模型。
与亚马逊自研模型的进展形成鲜明对比的是,其发展并未取得预期成果。
亚马逊CEO安迪·贾西在财报电话会上指出,自研模型并非取得成功的关键因素,行业不会出现单一AI模型一家独大的局面。
Jassy stated: “Even without developing frontier models, Amazon and AWS are still able to build a large-scale and extremely successful business.”
业内指出,这一表态标志着AI行业发展逻辑的重大转变。自2022年ChatGPT掀起全球AI热潮以来,“开发最强模型”一直被市场奉为圭臬,各大科技巨头纷纷砸下巨资投入技术研发与模型训练。
Amazon is now centered on a shift in the core of this competition from the pursuit of raw model performance alone to the more efficient deployment and operation of AI systems, seeking the so-called intelligence per dollar. Amazon holds an advantage over some competitors in this arena.
In the view of some media outlets, Amazon has never placed self-developed models at the highest priority in its development strategy.
Although the company had previously launched its Nova series of models, that series has remained unable to match frontier levels. Reports indicate that Amazon is gradually phasing out most of its existing Nova product line and redirecting resources to new frontier model projects.
Bedrock成战略核心
Jassy stated that clients are increasingly inclined to use multiple top-tier models simultaneously, and this trend serves as the primary driver for the explosive growth of the Amazon Bedrock platform.
Bedrock is a fully managed service and development platform for foundation models launched by Amazon Web Services, allowing customers to call the foundation models of multiple vendors through a unified API and develop generative AI applications.
业内指出,亚马逊的长期战略并非依赖自家模型打天下,而是将Bedrock打造成行业顶尖AI模型的一站式入口,无论这些模型是否由亚马逊亲生。
贾西表示:
如果一家企业要开发重要的AI应用,你肯定希望能够调用所有可用模型。各家模型会在不同阶段交替领先。
Jassy predicts that within the next few years at least five or six top-tier models of comparable performance will appear on the market.
Jassy stated: “They will all be launched on Bedrock and one of them will be our own.”
值得注意的是,亚马逊目前仍将持续投入自研AI模型。
Jassy pointed out that having a leading model not only enables Amazon to better control the costs of its own consumer applications and AWS clients but also allows the company to prioritize the functions valued by customers while reducing dependence on external vendors and accelerating the iteration process.
有媒体此前报道称,亚马逊的智能语音助手Alexa最近正在减少对Anthropic模型的依赖,改为采用自家研发的AI模型,以此降低运营成本。
来源:亚马逊亮眼业绩背后:跳出“最强模型”内卷,走不一样的AI突围之路 | 财联社