From: aidotengineer
Frontier Feud, a game show based on a survey of 100 AI Engineers, highlighted several aspects of AI industry trends, influential figures, and future considerations in AI development [00:01:07].
Influential AI Researchers
When 100 AI Engineers were asked to name the most influential AI researcher, the following names were among the top answers [00:05:27]:
- Jeff Hinton [00:06:21]
- Yan LaCun [00:07:22]
- Andrew Ng [00:09:09]
- Fei-Fei Li [00:09:13]
- Yoshua Bengio [00:09:18]
Considerations When Choosing an AI Model
The top considerations for AI Engineers when choosing a model were revealed as [00:09:41]:
- Cost: This was the number one surprising answer [00:10:47].
- Latency [00:11:04]
- Accuracy/Performance (eval benchmark scores) [00:11:21]
- Open Source vs. Closed Source [00:12:45]
- Capabilities [00:13:18]
AI Buzzwords Engineers are Tired Of
AI Engineers expressed fatigue with certain buzzwords [00:13:48]:
- AGI (Artificial General Intelligence): This was the number one answer [00:15:21].
- Agents [00:13:56]
- Co-pilots [00:15:12]
- RAG (Retrieval Augmented Generation) [00:16:21]
- Prompt Engineering [00:16:25]
AI Tools, Papers, and Challenges
Survey results also revealed common experiences and perspectives among AI Engineers [00:17:06]:
Favorite AI Tool
- Cursor was identified as an AI tool that engineers love, being the number one answer [00:17:34].
Job Most At Risk of AI Disruption
- Data Entry was cited as the job most at risk of AI disruption [00:20:43].
- Software Engineers were also mentioned as at risk [00:17:44].
Most Influential AI Paper
- “Attention Is All You Need” (the Transformer paper) was overwhelmingly cited as the most influential AI paper in history [00:17:47].
Biggest Nightmare for an AI Engineer
- An outage of the model was the top answer for an AI Engineer’s biggest nightmare at 2 AM [00:20:52].
- Other responses included “Hardware failure” [00:17:55] and “Cuda SS” [00:20:32].
Industry to Benefit Most from AI
- The Healthcare industry was identified as one that would benefit the most from AI [00:20:57].
Tech Hot Takes on the Future of AI
Participants shared their insights into the future of AI technology:
- Market Consolidation in Model Training: One prediction suggested that out of the current “fiveish people training big models today,” at least one will no longer be doing so by the end of the year [00:01:55].
- AI as Therapists: AIs are believed to make very good therapists [00:02:13].
- Importance of Human Thought: Engineers in the AI industry should dedicate time to “non-AI interrupted thoughts” to ensure good content for future AI models that represent users [00:02:25].
- On-Device Models and Specialization: Within a year and a half, the majority of deployed AI models are predicted to be on-device [00:03:31]. This trend favors smaller, orchestrated models that are hyperspecialized for specific tasks, rather than solely relying on larger models requiring external data transfer [00:03:40].
- AI in Dating: A futuristic view suggested that dating apps will evolve to have AIs dating each other [00:04:13].
- Beyond Transformers: The Transformer architecture may not be the final design for AI models, with the eventual development of models using biological materials [00:04:19].
- Multi-Bot Conversations: Future conversations are expected to involve at least two other bots in the room, moving beyond one-on-one interactions with a single bot [00:04:37].
State of AI Engineering Survey
A massive survey on the state of AI engineering is currently underway, aiming to provide deeper insights into the tools and workflows used by AI Engineers [00:21:39]. The results will be presented at the June AI Engineer World Fair, promoting industry transparency [00:21:42].