Fotis Savva

PhD Computer Science @UofGlasgow Looking into ways Machine Learning can expedite Analytics in Large Scale systems.

[Fotis Savva @ThinkinRandom]

@matsonj I really like the points under "What to actually build". Your "semantic layer" should be the DB . For enterprises which this might not be possible, a collection of previous (NLQ, SQL) and markdown documents that can be modified by the agent can still do the trick

[Fotis Savva @ThinkinRandom]

@GergelyOrosz What about all these companies saying AI replaced X% of their workforce or writes Y% percent of the code ?

[👩‍💻 Paige Bailey @DynamicWebPaige]

✨ @Microsoft's New Future of Work Report just dropped, and it's overwhelmingly focused on how AI impacts every aspect of information work. Adding some highlights below, but strongly recommend taking a look through the slides!

https://t.co/MmfBKRgMbV

📓 TL;DR - information workers complete simulated information work tasks much faster and with a higher quality of output when using generative AI based tools, with minimal impacts on quality. Experienced workers are less frustrated by mundane tasks, and inexperienced workers have their productivity accelerated.

Learning how to prompt effectively can be challenging. It is important to have strong user experience in the loop, to give users the ability to interrogate outputs and have some insight into uncertainty for outputs. For software engineering, the benefits of LLMs depend quite a bit on the task -- and AI is much more effective when breaking down large, complex tasks into more tractable subtasks.

There is a risk when using LLMs for education that students might use them as "steroids" rather than "coaches".

🚀 Nuggets from the slides:

• People took 37% less time on common writing tasks (Noy & Zhang 2023).
• BCG consultants produced >40% higher quality on one simulated consulting project (Dell’Acqua et al. 2023).
• Users were also 2x faster at solving simulated decision-making problems when using LLM-based search over traditional search (Spatharioti et al. 2023).

Study participants with Copilot completed experimenter-designed tasks in 26-73% as much time as those without it.
• A survey of enterprise users with access to Copilot also showed substantial perceived time savings.
• 73% agreed that Copilot helped them complete tasks faster, and 85% said it would help them get to a good first draft faster.
• Many studies found no statistically significant or meaningful effect on quality.

The study of M365 Defender Security Copilot found security novices with Copilot were 44% more accurate in answering questions about the security incidents they examined.

• Of enterprise Copilot users, 68% of respondents agreed that Copilot actually improved quality of their work.
• Users also reported tasks required less effort with Copilot. In the Teams Meeting Study, participants with access to Copilot found the task to be 58% less draining than
participants without access.
• Among enterprise Copilot users, 72% agreed that Copilot helped them spend less mental effort on mundane or repetitive tasks.

• In studying the staggered rollout of a generative AI-based conversational assistant, Brynjolfsson et al. (2023) found that the tool helped novice and low-skilled workers the most. They found suggestive evidence that the tool helped disseminate tacit knowledge that the experienced and high-skilled workers already had.

As AI is applied to more generative tasks, human work is shifting to “critical integration” of AI output, requiring expertise and judgement (Sarkar 2023). Moving beyond just error correction, AI provocateurs would challenge assumptions, encourage evaluation, and offer counterarguments.

The concept of “microproductivity”, in which complex tasks are decomposed into smaller subtasks and performed in “micromoments” by the person most skilled to do so, can be enhanced through automation (Teevan 2016). For example, Kokkalis et al. (2013) demonstrated that high level tasks broken into multistep action plans through crowdsourcing result in people completing significantly more tasks (47.1% task completion) compared to the control condition of no plans (37.8%). These benefits were scaled by applying NLP algorithms to automatically create action plans for a larger variety of tasks based on a training set of similar tasks, and the plans were further refined through human intervention.

[Fotis Savva @ThinkinRandom]

@Elliot_F Can you expand on this a little bit ?

[Fotis Savva @ThinkinRandom]

@jobergum Was this meant to provoke ? Why can't we have both ? I cannot see why the two need to be mutually exclusive

[MySQL @MySQL]

[Event] Analytics and Data Summit 2023, MAR 14-16, Redwood Shores, CA. Join us to learn about MySQL HeatWave - the One Database Service for OLTP, OLAP, Machine Learning - without ETL. https://t.co/Dw187U3l26
#AnDSummit23 #MySQL

[Fotis Savva @ThinkinRandom]

@michaelfreedman @mdwelsh Can you please elaborate a bit on this ? Is this a communication issue ? What if you restrict hiring remote to nearby time-zones ? I find that not offerring remote nowdays limits the worker pool

[Stefanos Sagkriotis @ssagkriotis]

Interested in In-network caching? This 9-post🧵 describes our approach to achieve "Scale-friendly In-network Coordination", accepted in GLOBECOM 22. https://t.co/j8bfd4A3Rr
@dpezaros @netlab_UofG

[Simon Willison @simonw]

Fascinating HN comment from someone who's company built a custom distributed data warehouse using compressed SQLite DB files in S3 that were queried using Lambda functions orchestrated by PostgreSQL running a custom foreign data wrapper https://t.co/gvRviN8D9R

[Shreya Shankar @sh_reya]

I probably should have written this years ago, but here are some MLOps principles I think every ML platform (codebase, data management platform) should have: 1/n

[Fotis Savva @ThinkinRandom]

@y0b1byte A PhD student actually spent time making this video instead of writing their dissertation 😂

[Andrea Baronchelli @a_baronca]

When you’re two years into your PhD and realize it really is *your* project.

[Arun Kumar @TweetAtAKK]

A fun aspect of working at the intersection of many areas--DB, ML/AI, systems, etc.-- is seeing how technical terms are wildly overloaded across areas of the same field (CS)! 😄 Some of my faves:

- Optimization
- Memory
- Network

What are your fave overloaded technical terms?

[Vasia Kalavri @vkalavri]

📢The SIGMOD 2022 Student Research Competition Call for Participation is out: https://t.co/eRbXG1eWJk
** Deadline: Nov 15 **
Please RT and encourage your students to submit an abstract! @2022Sigmod #sigmodsrc

[Eugene Yan @eugeneyan]

The first rule of machine learning: Start without machine learning

[Fotis Savva @ThinkinRandom]

@satnam6502 That's why I believe @GlasgowCS has different tracks now. But entry requirements are a completely different story. That being said you can become a great programmer without a CS degree

[UofG Computing Science @GlasgowCS]

Computing Science @UofGlasgow is offering up to 8 studentships to support #PhD research starting by January 2022. Info session for potential applicants will take place online on Wed 9 June at 12 noon https://t.co/SediCpMLwt

[Patrick Mineault @patrickmineault]

TIL Python classes can display custom HTML outputs in jupyter and colab via `_repr_html_`. That's how pandas table outputs work, for example. Here's Hello world and a map widget.

[Fotis Savva @ThinkinRandom]

"Tests are important but TDD is a damn cult." 😂

https://t.co/06opaHuDAu