Author: admin

More than 1,000 Amazon employees have signed an open letter expressing “serious concerns” about AI development, saying that the company’s “all-costs justified, warp speed” approach to the powerful technology will cause damage to “democracy, to our jobs, and to the earth.”

The letter, published on Wednesday, was signed by the Amazon workers anonymously, and comes a month after Amazon announced mass layoff plans as it increases adoption of AI in its operations.

Among the signatories are staffers in a range of positions, including engineers, product managers and warehouse associates.

Reflecting broader AI concerns across the industry, the letter was also supported by more than 2,400 workers from companies including Meta, Google, Apple and Microsoft.

The letter contains a range of demands for Amazon, concerning its impact on the workplace and the environment. Staffers are calling on the company to power all its data centers with clean energy, make sure its AI-powered products and services do not enable “violence, surveillance and mass deportation”, and form a working group comprised of non-managers “that will have significant ownership over org-level goals and how or if AI should be used in their orgs, how or if AI-related layoffs or headcount freezes are implemented, and how to mitigate or minimize the collateral effects of AI use, such as environmental impact”.

The letter was organized by employees affiliated with the advocacy group Amazon Employees for Climate Justice. One worker who was involved in drafting the letter explained that workers were compelled to speak out because of negative experiences with using AI tools in the workplace, as well as broader environmental concerns about the AI boom. The staffers, the employee said, wanted to advocate for a better way to develop, deploy and use the technology.

“I signed the letter because of leadership’s increasing emphasis on arbitrary productivity metrics and quotas, using AI as justification to push myself and my colleagues to work longer hours and push out more projects on tighter deadlines,” said a senior software engineer, who has been with the company for over a decade, and requested anonymity due to fear of reprisal.

Climate goals

The letter accuses Amazon of “casting aside its climate goals to build AI”.

Like other companies in the generative AI race, Amazon has invested heavily in building new data centers to power new tools – which are more resource intensive and demand high amounts of electricity to operate. The company plans to spend $150bn on data centers in the next 15 years, and just recently said it will invest $15bn to build data centers in northern Indiana and at least $3bn for data centers in Mississippi.

The letter claims that Amazon’s annual emissions have “grown roughly 35% since 2019”, despite the company’s promise in 2019 to achieve net zero carbon emissions by 2040. It warns many of Amazon’s investments in AI infrastructure will be in “locations where their energy demands will force utility companies to keep coal plans online or build new gas plants”.

“‘AI’ is being used as a magic word that is code for less worker power, hoarding of more resources, and making an uninformed gamble on high energy demand computer chips magically saving us from climate change,” said an Amazon customer researcher, who requested anonymity out of fear of retaliation for speaking out. “If we can build a climate saving AI – that’s awesome! But that’s not what Amazon is spending billions of dollars to develop. They are investing fossil fuel energy draining data centers for AI that is intended to surveil, exploit, and squeeze every extra cent out of customers, communities, and government agencies.”

In a statement to the Guardian, Amazon spokesperson Brad Glasser pushed back on employees’ claims and pointed toward the company’s climate goals. “Not only are we the leading data center operator in efficiency, we’re the world’s largest corporate purchaser of renewable energy for five consecutive years with over 600 projects globally,” said Glasser. “We’ve also invested significantly in nuclear energy through existing plants and new SMR technology–these aren’t distractions, they’re concrete actions demonstrating real progress toward our Climate Pledge commitment to reach net-zero carbon across our global operations by 2040.”

AI for productivity

The letter also includes strict demands around the role of AI in the Amazon workplace, demands that, staffers say, arose out of challenges employees are experiencing.

Three Amazon employees who spoke to the Guardian claimed that the company is pressuring them to use AI tools for productivity, in an effort to increase output. “I’m getting messaging from my direct manager and [from] of all the way up the chain, about how I should be using AI for coding, for writing, for basically all of my day-to-day tasks, and that those will make me more efficient, and also that if I don’t get on board and use them, that I’m going to fall behind, that it’s sort of sink or swim,” said a software engineer who has been with Amazon for over two years, requesting anonymity due to fear of reprisal.

The worker added that just weeks ago she was told by her manager that they were “expected to do twice as much work because of AI tools”, and expressed concern that the output expected demanded with fewer people is unsustainable, and “the tools are just not making up that gap.”

The customer researcher echoed similar concerns. “I have both personally felt the pressure to use AI in my role, and hear from so many of my colleagues they are under the same pressure …”.

“All the while, there’s no discussion about the immediate effects on us as workers – from unprecedented layoffs to unrealistic expectations for output.”

The senior software engineer said that the adoption of AI has had imperfect outcomes. He said that most commonly, workers are pressured to adopt agentic code generation tools: “Recently I worked on a project that was just cleaning up after a high-level engineer tried to use AI to generate code to complete a complex project,” said this worker. “But none of it worked and he didn’t understand why – starting from scratch would have actually been easier.”

Amazon did not respond to questions about the staffers’ workplace critiques about AI use.

Workers emphasized they are not against AI outright, rather they want it to be developed sustainably and with input from the people building and using it. “I see Amazon using AI to justify a power grab over community resources like water and energy, but also over its own workers, who are increasingly subject to surveillance, work speedups, and implicit threats of layoffs,” said the senior software engineer. “There is a culture of fear around openly discussing the drawbacks of AI at work, and one thing the letter is setting out to accomplish is to show our colleagues that many of us feel this way and that another path is possible.”

ACD is an evolutionary conserved mechanism used by stem and progenitor cells to generate cell diversity during development and regulate tissue homeostasis in the adult. CSCs, present in many human tumors, can divide asymmetrically to generate intratumoral heterogeneity or symmetrically to expand the tumor by self-renewal. Over the past 15 years, it has been suggested that dysregulation of the balance between symmetric and ACDs in CSCs, favoring symmetric divisions, can trigger tumor progression in different types of cancer (Bajaj et al., 2015; Chao et al., 2024; Li et al., 2022), including mammary tumors (Cicalese et al., 2009; Dey-Guha et al., 2011), GBM (Chen et al., 2014), oligodendrogliomas (Sugiarto et al., 2011; Daynac et al., 2018), colorectal cancer (Bu et al., 2013; Hwang et al., 2014), and hepatocellular carcinoma (Hwang et al., 2014). Thus, it is of great relevance to get a deeper insight into the network of regulators that control ACD, as well as the mechanisms by which they operate in this key process.

Drosophila neural stem cells or NBs have been used as a paradigm for many decades to study ACD (Homem and Knoblich, 2012). NBs divide asymmetrically to give rise to another self-renewing NB and a daughter cell that will start a differentiation process. Over all these years, a complex network of ACD regulators that tightly modulate this process has been characterized. For example, the so-called cell-fate determinants, including the Notch inhibitor Numb, accumulate asymmetrically at the basal pole of mitotic NBs and are exclusively segregated to one daughter cell, promoting in this cell a differentiation process. The asymmetric distribution of cell-fate determinants in the NB is, in turn, regulated by an intricate group of proteins asymmetrically located at the apical pole of mitotic NBs, generically known as the ‘apical complex’. This apical complex includes kinases (i.e., aPKC), small GTPases (i.e., Cdc42, Rap1), and Par proteins (i.e., Par-6, Par3), among others (Homem and Knoblich, 2012).

Given the potential relevance of ACD in CSCs, we decided to take advantage of all the knowledge accumulated in Drosophila about the network of modulators that control asymmetric NB division. As a first approach, we aimed to analyze whether the levels of human homologs of known Drosophila ACD regulators were altered in human tumors. Specifically, we centered on human GBM, as the presence of CSCs (GSCs) has been shown in this tumor. The microarray we interrogated with GBM patient samples had some limitations. For example, not all the human gene homologs of the Drosophila ACD regulators were present (i.e., the human homologs of the determinant Numb). Likewise, we only tested seven different GBM patient samples. Nevertheless, the output from this analysis was enough to determine that most of the human genes tested in the array presented altered levels of expression. We selected for further analyses RAP2A, one of the human genes that showed the lowest levels of expression compared to the control samples. However, it would be interesting to analyze in the future the potential consequences that altered levels of expression of the other human homologs in the array can have in the behavior of the GSCs. In silico analyses, taking advantage of the existence of established datasets, such as the TCGA, can help to more robustly assess, in a bigger sample size, the relevance of those human genes’ expression levels in GBM progression, as we observed for the gene RAP2A.

We have previously shown that Drosophila Rap1 acts as a novel NB ACD regulator in a complex with other small GTPases and the apical regulators Canoe (Cno), aPKC, and Par-6 (Carmena et al., 2011). Here, we have shown that Drosophila Rap2l also regulates NB ACD by ensuring the correct localization of the ACD modulators Cno and Numb. We have also demonstrated that RAP2A, the human homolog of Drosophila Rap2l, behaves as an ACD regulator in GBM neurosphere cultures, and its restitution to these GBM cultures, in which it is present at low levels, targets the stemness of GSCs, increasing the number of ACDs. It would be of great interest in the future to determine the specific mechanism by which Rap2l/RAP2A is regulating this process. One possibility is that, as it occurs in the case of the Drosophila ACD regulator Rap1, Rap2l/RAP2A is physically interacting or in a complex with other relevant ACD modulators. Thus, this study supports the relevance of ACD in CSCs of human tumors to refrain the expansion of the tumor. Other studies, however, claim that ACD should be targeted in human tumors as it promotes the intratumoral heterogeneity that hampers the complete tumor loss after chemotherapy (Samanta et al., 2023; Chao et al., 2023; Hitomi et al., 2021). More investigations should be carried out with other human gene homologs of ACD regulators to further confirm the results of this study. Likewise, analyses in vivo (i.e., in mouse xenografts) would also be required to reinforce our conclusions. This would be very relevant in order to consider ACD restitution in CSCs of human tumors, what has been called ‘differentiation therapy’ (de Thé, 2018), as a potential alternative therapeutic treatment.