Open AI heeft een krankzinnig fraaie uitbreiding (dit is mijn zeer gebalanceerde mening na een kwartiertje spelen) voor ChatGPT uitgebracht: Explore GPT’s. En daarmee kan je door een schier eindeloze lijst aan custom GPT’s bladeren of zoeken:

ChatGPT, custom GPT lijstscherm

Ik heb er een paar geprobeerd en Consensus (Your AI Research Assistant. Search 200M academic papers from Consensus, get science-based answers, and draft content with accurate citations) sprong er wat mij betreft bovenuit. Maar dat is geheel persoonlijk, want voor plaatjes van twee- tot zesvingerige gnomen is er natuurlijk ook van alles te vinden.

Ik heb aan Consensus de volgende vraag gesteld: What is the academic mood with respect to the upcoming European AI Act?

En dit kreeg ik terug van Consensus:

The academic mood regarding the upcoming European AI Act is diverse, with various studies highlighting different aspects of the proposed legislation:

1. Global Influence and Regulatory Standards: Gstrein (2022) notes that the European Commission’s AI Act proposal, following years of consultation and the work of the AI High Level Expert Group (AI HLEG), positions the EU to establish globally influential regulatory standards for AI. This Act could potentially integrate with existing and emerging legal frameworks, balancing the protection of human rights with technological development (Gstrein, 2022).

2. Metrics and Explainability: Sovrano, Sapienza, Palmirani, and Vitali (2022) discuss the AI Act’s focus on transparency, lawfulness, and fairness, emphasizing the importance of explainability metrics for high-risk AI systems. They suggest that compliance metrics should be risk-focused, model-agnostic, goal-aware, intelligible, and accessible (Sovrano et al., 2022).

3. Analysis of the Act’s Provisions: Veale and Zuiderveen Borgesius (2021) present an overview of the AI Act, analyzing its implications. They find that while some aspects, like different rules for varying risk levels of AI, are sensible, other provisions may have unexpected legal implications or might be ineffective. They raise concerns about the enforcement regime and the impact of maximum harmonisation on AI policy (Veale & Zuiderveen Borgesius, 2021).

4. Legal and HCI Implications: Urquhart, McGarry, and Crabtree (2022) explore the AI Act’s implications for Human-Computer Interaction (HCI) in designing trustworthy autonomous systems. They note the Act’s potential to regulate AI development stages, emphasizing design for human oversight and technical documentation requirements (Urquhart et al., 2022).

5. AI in Finance: Stiefmueller (2022) examines the AI Act’s impact on the financial sector, highlighting its balance between promoting AI uptake and addressing associated risks. The Act classifies certain financial applications as “high risk,” requiring specific requirements and supervision (Stiefmueller, 2022).

6. Impact on Product Safety and Liability: De Graaf and Veldt (2022) assess the AI Act’s relationship with existing product safety laws and its consequences for private law liability of AI system providers and users. They discuss the proposed risk-based approach and the implications for contractual relationships (De Graaf & Veldt, 2022).

In conclusion, the academic perspective on the European AI Act is multifaceted, addressing global regulatory influence, the importance of metrics and explainability, legal implications, the role of HCI, impacts on specific sectors like finance, and considerations around product safety and liability. These studies indicate a cautiously optimistic view, acknowledging the potential for positive impact while also highlighting areas of concern and improvement.

Het eerste wat mij opvalt is dat geen enkele van de referenties uit 2023 kwam; er is naar alle waarschijnlijkheid dus een cut-off ergens in 2022 met betrekking tot de gebruikte bronnen. Maar het is tegelijkertijd een prachtige eerste bron om uit te putten. Ik denk niet dat het de laatste keer is dat ik Consensus heb gebruikt…