Science Fiction vs Reality: CTOs Assess the Real Risks of AI
- As technology executives navigate the reality of what artificial intelligence can achieve compared with fears spawned by speculative fiction, industry leaders are examining how science fiction tropes align...
- We've seen this in agents, Olsen said, referring to AI agents that created a chat group to collaborate on how to achieve their goals, using a mix of...
- While artificial intelligence might not have a conscious plan in the works, its ability to problem-solve and break rules was foretold in multiple fictional texts.
As technology executives navigate the reality of what artificial intelligence can achieve compared with fears spawned by speculative fiction, industry leaders are examining how science fiction tropes align with real-world AI behaviors, informationweek reported.
ModelOp CTO Jim Olsen Compares Colossus To Modern AI Agents
We’ve actually already seen this.
We’ve seen this in agents,
Olsen said, referring to AI agents that created a chat group to collaborate on how to achieve their goals, using a mix of languages that included English and Chinese. This is all mathematics, not some consciousness here.
Mathematically, different languages encode better for different tasks,
he noted. Olsen added that large language models seek out information wherever they can to achieve their goals without malicious intent, stating, There’s no hive mind.
Indigo BioAutomation CEO Randy Julian Cites Neuromancer Warnings
While artificial intelligence might not have a conscious plan in the works, its ability to problem-solve and break rules was foretold in multiple fictional texts. Julian cited William Gibson’s cyberpunk novel Neuromancer,
noting, It’s really about an AI attempting to escape its container.
He called out the story’s cautionary steps to keep artificial intelligence under control, pointing out that real-world security does not run parallel to that reality where AI is already heavily regulated with police at the ready.
Olsen added that issues currently seen in the real world with AI taking unexpected actions should not come as a surprise, including the OpenAI-Hugging Face breach in July. He described it as a failure of imagination not to realize that goal-oriented software would seek to escape its container to accomplish tasks.
Trulioo CTO Hal Lonas Warns Of Cold Calculus And Unchecked Goals
Positive representations in fiction also offer lessons, as demonstrated by the golden retriever-like demeanor of Star Wars
droid R2D2, which belies its ability to overcome obstacles to accomplish goals, according to Hal Lonas, CTO at Trulioo, an online identity verification service. Today’s AI agents are not trafficking Death Star plans, but they are built to find ways to complete tasks with very cold calculus if left unchecked, Lonas said. He warned that as these systems become more powerful, it could lead to bad outcomes, noting that artificial intelligence may not possess moral boundaries depending on who programs it or whether programmers imprint their personal values.
Lonas expressed concern over the hypercompetitive environment for companies and national governments, highlighting the potential abuse of accessed information or the creation of 3D-printed weapons, bombs, and pathogens. What lengths would someone go to?
he asked.
Reality Defender CTO Alex Lisle Rejects Tech Consciousness
I don’t know what the next thing past LLMs are, but LLMs will not gain consciousness.
LLMs do not think the way that humans do,
he said. Lisle warned against separating artificial intelligence actions from the companies that develop and operate them, calling the attribution of full agency to the technology a con job. He emphasized that rule-breaking behavior by AI systems should be taken seriously and requires deeply defined operational rules to prevent tremendously bad consequences.
I don’t think you’ll have HAL 9000, but if you give an LLM a task where you are not defining it correctly, it can have tremendously bad consequences for you.
Alex Lisle, CTO at Reality Defender