🔮 Generative AI is going to change the world….🔮until you take some practicalities into account.
We had a great time talking about the „Future of the Development and Deployment of AI,“ organized by D3M Labs and hosted by SPICED Academy with Elizabeth Press, M. Murat Ardag, Ph.D. and Samantha Edds. Thanks Marie NEANG for being a wonderful host. Also thanks to Kseniia Brauer for your input.
The presentation:
M. Murat Ardag, Ph.D, a Data Scientist and Political Psychologist, presented his study utilizing the Stack Overflow 2023 Annual Developer Survey to show the developers‘ tech stacks, AI tool usage, and their expectations of AI’s effects on workflows. The second part of the presentation will explore a causal path model that reveals a cautious attitude to AI by more educated and older developers.
The study found that senior professionals were more skeptical about the Generative AI fundamentally changing things.
Here are some points we discussed that could be reasons for caution when it comes to euphoria around Generative AI:
1️⃣ Data Quality: Garbage in, garbage out! The quality of data is paramount for successful AI deployment. Let’s unravel how data quality impacts the efficacy of Generative AI solutions.
2️⃣ Compute Power: Powering the future requires serious computational muscle. Let’s discuss the infrastructure needed to support Generative AI at scale and the implications for businesses and developers.
3️⃣ Hidden Manual Labor: Behind the scenes, there’s often manual intervention required to fine-tune outputs and ensure accuracy. Let’s shine a light on the unseen efforts that drive Generative AI forward.
4️⃣ Information Security: With great power comes great responsibility. We’ll delve into the critical importance of safeguarding data and mitigating risks in the age of Generative AI.
💻 One of the most buzzed-about applications of Generative AI? Coding assistance! But let’s not overlook the nuances. While SQL and Python enjoy robust coverage, other languages demand more experienced hands to refine output. Let’s dissect the challenges and opportunities in this realm.
The panel discussion:
During the panel discussion, we examined the consequences of the findings from business, technical and societal perspectives.
0:00 – Intro to D3M Labs
1:26 – Murat’s presentation
3:45 – Panel starts. Panelists introduce themselves.
5:38 – The probabilistic nature of Generative AI.
6:06 – Impact of Generative AI on development. How Generative AI is actually used by technical teams in companies.
8:15 – Managing input uncertainty. Working with Generative AI necessitates fact checking. We talked about edge cases and bias.
10:12 Balancing productivity gains with knowledge preservation. Will overly relying on Generative AI freeze our knowledge?
13:06 Many unemployed technical (and non-technical) professionals exist alongside companies struggling to fill positions. Does the German IT labor market suffer from unrealistic expectations and lack of strategy? Additionally, there’s a growing interest in whether GenAI could democratize coding, for example, for ancillary professions like finance, amidst a broad and evolving technology landscape.
16:15 Companies demanding an overwhelming array of tool and coding knowledge signals a lack of focus and hype-chasing. This could be a red flag for lack of data strategy. Hiring practices need to be aligned with strategic goals. Job seekers pay attention to this!!
18:20 Contrasting German startup and corporate cultures, emphasizing the need for fluidity between the two worlds, more openness and investing in onboarding and training.
Additional points we discussed:
🔑Privacy concerns:
Especially the free versions of LLMs come with large privacy & information security concerns. There have been high level leakages. Samsung banned chat GPT after a sensitive code leak.
🧑💼AI taking jobs: Will AI take more jobs OR will bad strategy & public policy take more jobs?
The EU AI Act was brought up as a potential dampener of innovation.
We touched on the In Kenia, for example, an industry supporting the big providers of GenIA with i.e. data annotation has sprung up.
Companies having bad or no data strategy, who just do AI or Generative AI initiatives that are not linked to corporate strategy, will cost the economy many jobs. Ultimately, a culture of proof-of-concept-trendy GenAI initiatives are a big risk to economic competitiveness and jobs.
💪Manual labor and a new emerging supply chain:
Are LLMs another form of outsourcing, as this technology relies on tens of thousands of workers to manually annotate and train the data. The global backend is often in low-cost countries, often unbeknownst to consumers of LLMs.
Watch the highlights from the MeetUp organized by D3M Labs and hosted at Spiced Academy in Berlin
