Decision Management Community

Today Google announced Gemini 1.5 that supports “millions of tokens of multimodal input. The multimodal capabilities of the model means you can interact in sophisticated ways with entire books, very long document collections, codebases of hundreds of thousands of lines across hundreds of files, full movies, entire podcast series, and more”. Report Demo

Prof. Warren Powell: “Sorry, this is simply laughable. I would file this alongside Geoffrey Hinton’s prediction (circa 2016) that we will not need radiologists in 5 years (didn’t happen). Don’t these people ever learn? LLMs today certainly are useful to programmers, but only for filling in boilerplate code which can be learned from existing code. Software requires the creative guidance of a programmer to specify what task is being solved. Even with five more years of development, LLMs are never going to be able to guess what a piece of software needs to do, and my guess is that programming languages and tools will continue to evolve. My guess is that LLMs will make programmers more productive. You would think this might translate into fewer programmers, but I wouldn’t bet on this – I think we will use the same number of programmers to do more.” Link

It’s scary to think that LLMs could have embedded malicious sleeper agents. But a recent paper by Anthropic has been causing quite a stir online – they have proven that LLMs can have malicious “sleeper” behavior secretly embedded by a bad actor! The worst part of this is that this behavior cannot be detected or removed later. In one of their experiments, they trained models that would write good secure code if the year was 2023, but write exploitable code if the year was 2024. Does this finding reinforce the need for any company using LLMs to have human-in-the loop? Link