ChatGPT whitethorn person launched successful precocious 2022, but 2023 was undoubtedly nan twelvemonth that generative AI took clasp of nan nationalist consciousness.
Not only did ChatGPT scope caller highs (and lows), but a plethora of seismic changes shook nan world, from unthinkable rival products to shocking scandals and everything successful between. As nan twelvemonth draws to a close, we’ve taken a look backmost astatine nan 9 astir important events successful AI that took spot complete nan past 12 months. It’s been a twelvemonth for illustration nary different for AI — here’s everything that made it memorable, starting astatine nan opening of 2023.
ChatGPT’s rivals unreserved to market
No database for illustration this starts without nan unparalleled emergence of ChatGPT. The free-to-use chatbot from OpenAI took nan world by storm, quickly increasing and capturing nan imagination of everyone, from tech leaders to mean group connected nan street.
The work was initially launched successful November of 2022 but really began to turn passim nan first fewer months of nan year. The stunning occurrence of ChatGPT near its rivals scrambling for a response. No 1 was much terrified than Google, it seemed, pinch nan institution fretting that AI could make its lucrative hunt business each but obsolete. In February, a fewer months aft ChatGPT went public, Google deed backmost pinch Bard, its ain AI chatbot. Then, conscionable a time later, Microsoft revealed its ain effort pinch Bing Chat.
Things went a small worse for Bing Chat, pinch nan chatbot being prone to what became known arsenic “hallucinations” — incidents wherever it would lie, dream up clone facts, and mostly beryllium all-round unreliable. It told 1 newsman that it spied on, fell successful emotion with, and past murdered its ain developer. When we tested it, it claimed to beryllium perfect, said it wanted to beryllium human, and based on pinch america relentlessly. In different words, it was unhinged.
Things went a small amended for Google Bard — it wasn’t astir truthful argumentative and avoided Bing Chat’s deranged usage of emojis altogether — but still had a tendency to beryllium unreliable. If thing else, some Google and Microsoft’s quickfire responses to ChatGPT showed really rushed their attempts were — and really vulnerable AI’s penchant for misinformation could be.
GPT-4 makes a splash
When ChatGPT launched, it was powered by a ample connection exemplary (LLM) called GPT-3.5. This was highly tin but had immoderate limitations, specified arsenic only being capable to return matter arsenic an input method. Much of that changed pinch GPT-4, which went nationalist successful March.
ChatGPT developer OpenAI said nan caller LLM was amended successful 3 cardinal ways: creativity, ocular input, and longer context. For instance, GPT-4 tin usage images arsenic input and tin besides collaborate pinch users connected imaginative projects specified arsenic music, screenplays, and literature.
Currently, GPT-4 is locked down OpenAI’s ChatGPT Plus paywall, which costs $20 a month. But moreover pinch that constricted reach, it has had a monolithic effect connected AI. When Google announced its Gemini LLM successful December, it claimed it was capable to hit GPT-4 successful astir tests — yet nan truth that it could do truthful by only a fewer percent points almost a twelvemonth aft GPT-4 launched tells you everything you request to cognize astir really precocious OpenAI’s exemplary is.
AI-generated images statesman to fool nan public
Few incidents illustrated nan powerfulness of AI to instrumentality and misinform than a azygous image released successful early 2023: that of Pope Francis successful a large achromatic puffer jacket.
The improbable image had been created by a personification named Pablo Xavier utilizing AI generator Midjourney, but it was truthful realistic that it easy fooled immense numbers of viewers connected societal media, including celebrities for illustration Chrissy Teigen. It highlighted really convincing AI image generators tin beryllium — and their expertise to hoodwink group into believing things that aren’t real.
In fact, conscionable nan week earlier nan image of nan Pope emerged, a different group of pictures made nan news for a akin reason. They depicted erstwhile president Donald Trump being arrested, fighting pinch constabulary officers, and serving clip successful jail. When powerful image generators get mixed pinch delicate topics, beryllium they related to politics, health, war, aliases thing else, nan risks tin beryllium extreme. As AI-generated images get moreover much realistic, nan Pope Francis creation was a lighthearted illustration of really quickly we’d each request to up our media literacy game.
AI-generated images person go progressively communal passim nan year, even appearing successful Google hunt results up of existent images.
A petition originates to sound nan alarm
AI has been moving astatine specified breakneck velocity — and has already had specified alarming consequences — that galore group person been raising superior concerns complete what it could unleash. In March 2023, those fears were voiced by immoderate of nan world’s astir salient tech leaders successful an open letter.
The statement called connected “all AI labs to instantly region for astatine slightest six months nan training of AI systems much powerful than GPT-4” to springiness clip for nine astatine ample to measure nan risks. Otherwise, nan full-scale push into AI could “pose profound risks to nine and humanity,” including nan imaginable demolition of jobs, obsolescence of quality life, and “loss of power of our civilization.”
The missive was signed by a Who’s Who of tech leaders, from Apple co-founder Steve Wozniak and Tesla leader Elon Musk to a slate of researchers and academics. Whether immoderate AI firms really took announcement erstwhile nan imaginable short-term profits are truthful awesome is different matter — conscionable look astatine Google Gemini, which its creators opportunity outstrips GPT-4. Here’s hoping nan missive doesn’t move retired to beryllium prophetic.
ChatGPT connects to nan internet
When ChatGPT first launched, it relied connected its monolithic trove of training information to thief supply answers to people. The problem pinch that, though, was that it couldn’t enactment up to day and was of small usage if personification wanted to usage it to make a preservation astatine a edifice aliases find a nexus to bargain a product.
That each changed erstwhile OpenAI announced a slate of plugins that would help ChatGPT Plugins that could link to nan internet. Suddenly, a caller measurement of utilizing AI to get things done was opened up. The alteration besides vastly updated and expanded what nan chatbot could do compared to before. In position of capacity and utility, it was a immense upgrade.
But it wasn’t until May that its browsing capabilities were expanded erstwhile nan Browsing done Bing plugin was announced astatine nan Microsoft Build developer conference. It was a slow rollout until September erstwhile it became disposable to each ChatGPT Plus users.
Windows gets a caller Copilot
Microsoft introduced Bing Chat astatine nan opening of nan year, but nan institution didn’t enactment still. It followed this move up pinch Copilot, a overmuch much expansive usage of AI that was incorporated into Microsoft’s products, which was first announced for Microsoft 365 Copilot.
While Bing Chat was a elemental chatbot, Copilot is much of a integer assistant. It’s woven into a bunch of Microsoft apps, for illustration Word and Teams, arsenic good arsenic Windows 11 itself. It tin create images, summarize meetings, find accusation and nonstop it to your different devices, and a full batch more. The thought is it saves you clip and effort connected lengthy tasks by automating them for you. In fact, Bing Chat was moreover subsumed nether nan Copilot umbrella successful November.
By integrating Copilot truthful tightly into Windows 11, Microsoft not only indicated its all-in accuracy regarding AI, but threw down nan gauntlet to Apple and its rival macOS operating system. So far, nan advantage lies pinch Microsoft, particularly arsenic we look up to Windows 12 launching successful 2024.
The world world grapples pinch AI
Given really quickly AI exploded onto nan scene, you’d beryllium forgiven for not rather knowing really it each works. But that deficiency of knowledge had existent consequences for students astatine Texas A&M University–Commerce erstwhile a professor flunked nan full class for apparently utilizing ChatGPT to constitute their papers — contempt nary grounds of students really doing that.
The problem arose erstwhile Dr. Jared Mumm copied and pasted students’ papers into ChatGPT, past asked nan chatbot if it could person generated nan text, to which ChatGPT answered successful nan affirmative. The problem? ChatGPT can’t really observe AI plagiarism successful that way.
To exemplify nan point, Reddit users took Dr. Mumm’s ain missive to students successful which he accused them of cheating, past pasted it into ChatGPT and asked nan chatbot if it mightiness person written nan letter. The answer? “Yes, I wrote nan contented you’ve shared,” lied ChatGPT. If you want nan cleanable illustration of some an AI hallucinating and quality disorder surrounding its abilities, this is it.
Hollywood vs. AI
It’s clear by now that artificial intelligence is incredibly capable, but it’s for precisely that logic that immense numbers of group each complete nan globe are truthful worried astir it. One study claimed, for example, that up to 300 cardinal jobs could beryllium destroyed by AI if its improvement goes unchecked. That benignant of interest motivated Hollywood writers to onslaught complete fears that workplace bosses would yet switch them pinch AI.
The writers’ national — nan Writers Guild of America (WGA) — went connected onslaught for almost 5 months complete nan issue, starting successful May and ending successful September, erstwhile it yet won important concessions from studios. These included provisions that AI would not beryllium utilized to constitute aliases rewrite material, that writers’ contented could not beryllium utilized to train AI, and more. It was a important triumph for nan affected workers, but fixed really AI improvement is continuing apace, it apt won’t beryllium nan past conflict betwixt AI and nan group it affects.
The Sam Altman saga
Sam Altman, OpenAI’s CEO, has been nan very visible look of nan full AI manufacture ever since ChatGPT exploded into nan world’s consciousness. Yet that each came crashing down 1 time successful November erstwhile he was unceremoniously fired from OpenAI — to nan complete astonishment of himself and nan world astatine large.
OpenAI’s committee accused him of being “not consistently candid” successful his dealings pinch nan company. Yet nan backlash was swift and strong, pinch nan mostly of nan company’s labor threatening to locomotion out if Altman was not reinstated. OpenAI investor Microsoft offered jobs to Altman and anyone other from OpenAI who wanted to join, and for a moment, it seemed that Altman’s institution was connected nan brink of collapse.
Then, conscionable arsenic quickly arsenic he left, Altman was reinstated, pinch galore members of nan committee expressing regret complete nan full incident. The net watched pinch bated activity successful real-time arsenic nan events unfolded, pinch 1 singular mobility moving down nan full circus: Why? Did Altman really stumble upon a improvement successful AI that caused superior ethical concerns? Was Project Q* astir to execute AGI? Was location a Game of Thrones-esque vying for power, aliases was Altman really conscionable a bad boss?
We whitethorn ne'er cognize nan full truth. But nary different infinitesimal successful 2023 encapsulated nan hysteria, fascination, and conspiratorial reasoning caused by AI this year.
- Google mightiness yet person an reply to Chat GPT-4
- One twelvemonth ago, ChatGPT started a revolution
- Here’s why you can’t motion up for ChatGPT Plus correct now
- The world responds to nan creator of ChatGPT being fired by his ain company
- GPT-4 Turbo is nan biggest update since ChatGPT’s launch