data subtldr week 25 year 2024

The Reddit thread titled Sounds familiar? in the subreddit 't5_36en4' showcases the struggles of data engineers dealing with impractical or unorganized data requests. Key highlights include users sharing their experiences of being asked to pull complex data in an unreasonable timeframe, to run 'A.I. and machine learning' on simple Excel tasks, and to manage disorganized CRM systems. Some users express frustrations about their colleagues' inability to specify the data they need or use available systems effectively. The overall sentiment suggests a considerable level of exasperation among data engineers due to unrealistic demands and lack of understanding from non-technical staff. Humor is employed as a coping mechanism.

The Reddit thread discusses issues faced by users after the release of NumPy 2.0. Several users experienced failures in their pipelines, including unexpected errors even with non-direct dependencies like GeoPandas. Commenters suggested strategies to mitigate such issues, including pinning dependencies, using a lockfile, and using tools like poetry, rye, or flit. Also, it was emphasized that upgrades should not break production. Some users criticized the complexity of working with Python on Snowflake and the syntax of NumPy. Another user shared a solution for Snowflake users to pin the NumPy version within UDFs and Stored Procedures. Overall, the sentiment was mixed, with users expressing frustration but also sharing solutions.

The Reddit thread centers around a post highlighting the unrealistic expectations of job descriptions, using the example of requiring 10 years of experience in dbt, a software that's only been publicly available since 2016. The top comments add further context; deal_damage notes that the first stable version of dbt was released in 2021. Other users criticize the unrealistic expectations of job requirements, with dumbasfuck6969 sarcastically mentioning the expectation of technical skills combined with sales experience. They also express concern about the use of AI filters in job applications and worry that such requirements may be used to justify hiring non-local candidates.

The Reddit thread discusses the launch of Safe Superintelligence Inc. by Ilya Sutskever and his team. The top comments reflect skepticism towards the company's ambition to build Advanced Super Intelligence (ASI) without a clear roadmap or profit plan. Some users question the feasibility of such a project, jesting that the company might be better off as a tax-exempt religious institution. Others express respect for the team's AI expertise but still doubt the feasibility of the project. A few comments also humorously remark on the use of the term safe in the company's name, comparing it to the term open in OpenAI. Overall, the sentiment leans towards intrigue mixed with skepticism.

The Reddit thread revolves around a user seeking advice on their machine learning (ML) career. The user expresses feeling lost due to struggles with depression, ADHD, and imposter syndrome, despite holding a PhD and a managerial role in AI. Top comments suggest that the user should leverage their strengths in management and leadership rather than diving deeper into technical ML skills. They advise that a great manager is one who can effectively utilize their team's knowledge and put the right people in the right positions. Many comments also address imposter syndrome, suggesting the user might be suffering from it and encouraging them to trust themselves and their abilities. Overall, the sentiment is positive and supportive.

The Reddit thread discusses the challenges faced by industry-based Machine Learning researchers in finding time to publish papers. The top comments suggest that publishing is often part of the job description, and many are working more than the standard 40 hours per week. Some believe that certain jobs can implicitly or explicitly hinder one's research career, suggesting the need for an exit plan. The competitive nature of the field and the pressure from future employers about publishing papers were also highlighted. A few commenters noted that some workplaces encourage publication and provide support, such as facilitating conference attendance for employees with accepted papers. However, others felt that research does not respect personal time and is difficult to accomplish within a 9-5 schedule.