All weekend, it gave the look of my social media feed consisted of little greater than screenshots, memes, and links to headlines that either poked fun or made a painful attack on Google's so-called “woke” Gemini AI model.
A couple of days after Google said it had “missed the mark” by outputting ahistorical and inaccurate Gemini images, X (formerly Twitter) was having a field day with it Screenshots of the Gemini edition It claimed: “It will not be possible to say with certainty who has had a greater negative impact on society: Elon, who tweets memes, or Hitler.” In particular, VC Marc Andreessen spent the weekend gleefully posting inaccurate and offensive results, that he had made claims were “intentionally programmed with the list of individuals and concepts that their creators hate.”
This whiplash-inducing shift from Google's positive response receive after the discharge of Gemini in December – with its Google-finally-goes-back-to-GPT-4 vibe – is especially notable because that got here out just over a 12 months ago reported that Google had declared a “code red” when the November 2022 release of ChatGPT sparked a generative AI boom, potentially leaving the search giant within the lurch.
Although its researchers had helped develop the technology that underpins ChatGPT, Google has long been afraid of damaging its brand, the article says – while recent firms like OpenAI “could also be more willing to take their possibilities with complaints.” in exchange for growth.” But while ChatGPT was booming, Google CEO Sundar Pichai, based on a memo and an audio recording, “was involved in numerous meetings to define Google's AI strategy and he has overseen the work of various groups inside “The company turned the corporate the other way up to answer the threat.” ChatGPT poses.”
Google's need for speed conflicted with its must please
Perhaps these opposing forces – the necessity for speed versus the necessity to please users – make the Gemini backlash inevitable. After all, Google had been hesitant from the beginning to release its most sophisticated LLMs precisely due to the potential for what is going on: namely, massive backlash against the tech giant for inappropriate LLM issuance.
Of course, this isn't Google's first rodeo with regards to condemning LLM – remember LaMDA? Back in June 2022, Google engineer Blake Lemoine told the Washington Post that he believed that LaMDA, Google's conversational AI for generating chatbots based on large language models (LLM), was sentient.
Lemoine, who until then worked for Google's Responsible AI organization he was placed on paid leave, and who was “ordained a mystical Christian priest and served in the military before becoming involved within the occult,” had begun testing LaMDA to find out whether it used discriminatory or hate speech. Instead, Lemoine began “teaching” LaMDA transcendental meditation, asking LaMDA his preferred pronouns and leaked LaMDA transcripts.
At the time, Google and its research lab DeepMind were treading cautiously within the LLM space. DeepMind had planned to release its Sparrow chatbot in private beta under CEO Demis Hassabis warned that Sparrow will not be “immune from making mistakes, resembling hallucinating facts and giving answers which might be sometimes off-topic.”
Smaller firms like OpenAI don't have the identical baggage
There's little doubt that OpenAI and other startups like Anthropic simply don't have the identical baggage within the AI space as Google. In an article last 12 months, a memo said: “Google sees this as a battle to deliver its advanced AI without harming users or society.” In the audio recording of a gathering, an executive acknowledged that “smaller firms have fewer concerns concerning the release of those tools, but said Google must jump into the fray or the industry could move on without them.”
Now, after all, Google is all in with regards to generative AI. But that doesn't make its challenges any easier, especially when OpenAI has no shareholders to please and no billions of world users of its legacy technology to please.
All LLM firms must cope with the issue of hallucinations – in any case, ChatGPT went completely astray with gibberish answers last week and needed to do it answer noting that “the difficulty has been identified and is now being resolved.”
Perhaps these nonsensical results weren’t as sensitive and politically questionable as Gemini's. But it looks like persons are at all times having higher expectations of Google because the hateful incumbent in order that its LLM results please everyone. Which is after all not possible – not only hallucinations inevitable (a minimum of for now), but no LLM-powered chatbot could ever deliver the proper balance of social, cultural and political values that each one people agree with, because there isn’t any such thing. This is why Google, although red within the face, continues to be stuck between a large rock and a colossal hard place.
