a.heateor_sss_amp{padding:0 4px}div.heateor_sss_horizontal_sharing a amp-img{display:inline-block}.heateor_sss_amp_instagram img{background-color:#624E47}.heateor_sss_amp_yummly img{background-color:#E16120}.heateor_sss_amp_youtube img{background-color:#ff0000}.heateor_sss_amp_buffer img{background-color:#000}.heateor_sss_amp_delicious img{background-color:#53BEEE}.heateor_sss_amp_facebook img{background-color:#3C589A}.heateor_sss_amp_digg img{background-color:#006094}.heateor_sss_amp_email img{background-color:#649A3F}.heateor_sss_amp_float_it img{background-color:#53BEEE}.heateor_sss_amp_linkedin img{background-color:#0077B5}.heateor_sss_amp_pinterest img{background-color:#CC2329}.heateor_sss_amp_print img{background-color:#FD6500}.heateor_sss_amp_reddit img{background-color:#FF5700}.heateor_sss_amp_stocktwits img{background-color:#40576F}.heateor_sss_amp_mewe img{background-color:#007da1}.heateor_sss_amp_mix img{background-color:#ff8226}.heateor_sss_amp_tumblr img{background-color:#29435D}.heateor_sss_amp_twitter img{background-color:#55acee}.heateor_sss_amp_vkontakte img{background-color:#5E84AC}.heateor_sss_amp_yahoo img{background-color:#8F03CC}.heateor_sss_amp_xing img{background-color:#00797D}.heateor_sss_amp_instagram img{background-color:#527FA4}.heateor_sss_amp_whatsapp img{background-color:#55EB4C}.heateor_sss_amp_aim img{background-color:#10ff00}.heateor_sss_amp_amazon_wish_list img{background-color:#ffe000}.heateor_sss_amp_aol_mail img{background-color:#2A2A2A}.heateor_sss_amp_app_net img{background-color:#5D5D5D}.heateor_sss_amp_baidu img{background-color:#2319DC}.heateor_sss_amp_balatarin img{background-color:#fff}.heateor_sss_amp_bibsonomy img{background-color:#000}.heateor_sss_amp_bitty_browser img{background-color:#EFEFEF}.heateor_sss_amp_blinklist img{background-color:#3D3C3B}.heateor_sss_amp_blogger_post img{background-color:#FDA352}.heateor_sss_amp_blogmarks img{background-color:#535353}.heateor_sss_amp_bookmarks_fr img{background-color:#E8EAD4}.heateor_sss_amp_box_net img{background-color:#1A74B0}.heateor_sss_amp_buddymarks img{background-color:#ffd400}.heateor_sss_amp_care2_news img{background-color:#6EB43F}.heateor_sss_amp_citeulike img{background-color:#2781CD}.heateor_sss_amp_comment img{background-color:#444}.heateor_sss_amp_diary_ru img{background-color:#E8D8C6}.heateor_sss_amp_diaspora img{background-color:#2E3436}.heateor_sss_amp_dihitt img{background-color:#FF6300}.heateor_sss_amp_diigo img{background-color:#4A8BCA}.heateor_sss_amp_douban img{background-color:#497700}.heateor_sss_amp_draugiem img{background-color:#ffad66}.heateor_sss_amp_dzone img{background-color:#fff088}.heateor_sss_amp_evernote img{background-color:#8BE056}.heateor_sss_amp_facebook_messenger img{background-color:#0084FF}.heateor_sss_amp_fark img{background-color:#555}.heateor_sss_amp_fintel img{background-color:#087515}.heateor_sss_amp_flipboard img{background-color:#CC0000}.heateor_sss_amp_folkd img{background-color:#0F70B2}.heateor_sss_amp_google_classroom img{background-color:#FFC112}.heateor_sss_amp_google_bookmarks img{background-color:#CB0909}.heateor_sss_amp_google_gmail img{background-color:#E5E5E5}.heateor_sss_amp_hacker_news img{background-color:#F60}.heateor_sss_amp_hatena img{background-color:#00A6DB}.heateor_sss_amp_instapaper img{background-color:#EDEDED}.heateor_sss_amp_jamespot img{background-color:#FF9E2C}.heateor_sss_amp_kakao img{background-color:#FCB700}.heateor_sss_amp_kik img{background-color:#2A2A2A}.heateor_sss_amp_kindle_it img{background-color:#2A2A2A}.heateor_sss_amp_known img{background-color:#fff101}.heateor_sss_amp_line img{background-color:#00C300}.heateor_sss_amp_livejournal img{background-color:#EDEDED}.heateor_sss_amp_mail_ru img{background-color:#356FAC}.heateor_sss_amp_mendeley img{background-color:#A70805}.heateor_sss_amp_meneame img{background-color:#FF7D12}.heateor_sss_amp_mixi img{background-color:#EDEDED}.heateor_sss_amp_myspace img{background-color:#2A2A2A}.heateor_sss_amp_netlog img{background-color:#2A2A2A}.heateor_sss_amp_netvouz img{background-color:#c0ff00}.heateor_sss_amp_newsvine img{background-color:#055D00}.heateor_sss_amp_nujij img{background-color:#D40000}.heateor_sss_amp_odnoklassniki img{background-color:#F2720C}.heateor_sss_amp_oknotizie img{background-color:#fdff88}.heateor_sss_amp_outlook_com img{background-color:#0072C6}.heateor_sss_amp_papaly img{background-color:#3AC0F6}.heateor_sss_amp_pinboard img{background-color:#1341DE}.heateor_sss_amp_plurk img{background-color:#CF682F}.heateor_sss_amp_pocket img{background-color:#f0f0f0}.heateor_sss_amp_polyvore img{background-color:#2A2A2A}.heateor_sss_amp_printfriendly img{background-color:#61D1D5}.heateor_sss_amp_protopage_bookmarks img{background-color:#413FFF}.heateor_sss_amp_pusha img{background-color:#0072B8}.heateor_sss_amp_qzone img{background-color:#2B82D9}.heateor_sss_amp_refind img{background-color:#1492ef}.heateor_sss_amp_rediff_mypage img{background-color:#D20000}.heateor_sss_amp_renren img{background-color:#005EAC}.heateor_sss_amp_segnalo img{background-color:#fdff88}.heateor_sss_amp_sina_weibo img{background-color:#ff0}.heateor_sss_amp_sitejot img{background-color:#ffc800}.heateor_sss_amp_skype img{background-color:#00AFF0}.heateor_sss_amp_sms img{background-color:#6ebe45}.heateor_sss_amp_slashdot img{background-color:#004242}.heateor_sss_amp_stumpedia img{background-color:#EDEDED}.heateor_sss_amp_svejo img{background-color:#fa7aa3}.heateor_sss_amp_symbaloo_feeds img{background-color:#6DA8F7}.heateor_sss_amp_telegram img{background-color:#3DA5f1}.heateor_sss_amp_trello img{background-color:#1189CE}.heateor_sss_amp_tuenti img{background-color:#0075C9}.heateor_sss_amp_twiddla img{background-color:#EDEDED}.heateor_sss_amp_typepad_post img{background-color:#2A2A2A}.heateor_sss_amp_viadeo img{background-color:#2A2A2A}.heateor_sss_amp_viber img{background-color:#8B628F}.heateor_sss_amp_wanelo img{background-color:#fff}.heateor_sss_amp_webnews img{background-color:#CC2512}.heateor_sss_amp_wordpress img{background-color:#464646}.heateor_sss_amp_wykop img{background-color:#367DA9}.heateor_sss_amp_yahoo_mail img{background-color:#400090}.heateor_sss_amp_yahoo_messenger img{background-color:#400090}.heateor_sss_amp_yoolink img{background-color:#A2C538}.heateor_sss_amp_youmob img{background-color:#3B599D}.heateor_sss_amp_gentlereader img{background-color:#46aecf}.heateor_sss_amp_threema img{background-color:#2A2A2A}.heateor_sss_vertical_sharing{position:fixed;left:11px;z-index:99999}.heateor-total-share-count .sss_share_count{color:#666;font-size:23px}.heateor-total-share-count .sss_share_lbl{color:#666}.amp-wp-enforced-sizes img[alt="Pinterest"]{background:#cc2329}.amp-wp-enforced-sizes img[alt="Viber"]{background:#8b628f}.amp-wp-enforced-sizes img[alt="Print"]{background:#fd6500}.amp-wp-enforced-sizes img[alt="Threema"]{background:#2a2a2a}.amp-wp-article-content .heateor_sss_vertical_sharing{left:5px}.amp-wp-article-content amp-img[alt="Pinterest"]{left:4px}.amp-wp-enforced-sizes img[alt="MySpace"]{background:#2a2a2a} amp-web-push-widget button.amp-subscribe { display: inline-flex; align-items: center; border-radius: 5px; border: 0; box-sizing: border-box; margin: 0; padding: 10px 15px; cursor: pointer; outline: none; font-size: 15px; font-weight: 500; background: #4A90E2; margin-top: 7px; color: white; box-shadow: 0 1px 1px 0 rgba(0, 0, 0, 0.5); -webkit-tap-highlight-color: rgba(0, 0, 0, 0); } a.heateor_sss_amp{padding:0 4px;}div.heateor_sss_horizontal_sharing a amp-img{display:inline-block;}.heateor_sss_amp_gab img{background-color:#25CC80}.heateor_sss_amp_parler img{background-color:#892E5E}.heateor_sss_amp_gettr img{background-color:#E50000}.heateor_sss_amp_instagram img{background-color:#624E47}.heateor_sss_amp_yummly img{background-color:#E16120}.heateor_sss_amp_youtube img{background-color:#ff0000}.heateor_sss_amp_teams img{background-color:#5059c9}.heateor_sss_amp_google_translate img{background-color:#528ff5}.heateor_sss_amp_x img{background-color:#2a2a2a}.heateor_sss_amp_rutube img{background-color:#14191f}.heateor_sss_amp_buffer img{background-color:#000}.heateor_sss_amp_delicious img{background-color:#53BEEE}.heateor_sss_amp_rss img{background-color:#e3702d}.heateor_sss_amp_facebook img{background-color:#0765FE}.heateor_sss_amp_digg img{background-color:#006094}.heateor_sss_amp_email img{background-color:#649A3F}.heateor_sss_amp_float_it img{background-color:#53BEEE}.heateor_sss_amp_linkedin img{background-color:#0077B5}.heateor_sss_amp_pinterest img{background-color:#CC2329}.heateor_sss_amp_print img{background-color:#FD6500}.heateor_sss_amp_reddit img{background-color:#FF5700}.heateor_sss_amp_mastodon img{background-color:#6364FF}.heateor_sss_amp_stocktwits img{background-color: #40576F}.heateor_sss_amp_mewe img{background-color:#007da1}.heateor_sss_amp_mix img{background-color:#ff8226}.heateor_sss_amp_tumblr img{background-color:#29435D}.heateor_sss_amp_twitter img{background-color:#55acee}.heateor_sss_amp_vkontakte img{background-color:#0077FF}.heateor_sss_amp_yahoo img{background-color:#8F03CC}.heateor_sss_amp_xing img{background-color:#00797D}.heateor_sss_amp_instagram img{background-color:#527FA4}.heateor_sss_amp_whatsapp img{background-color:#55EB4C}.heateor_sss_amp_aim img{background-color: #10ff00}.heateor_sss_amp_amazon_wish_list img{background-color: #ffe000}.heateor_sss_amp_aol_mail img{background-color: #2A2A2A}.heateor_sss_amp_app_net img{background-color: #5D5D5D}.heateor_sss_amp_balatarin img{background-color: #fff}.heateor_sss_amp_bibsonomy img{background-color: #000}.heateor_sss_amp_bitty_browser img{background-color: #EFEFEF}.heateor_sss_amp_blinklist img{background-color: #3D3C3B}.heateor_sss_amp_blogger_post img{background-color: #FDA352}.heateor_sss_amp_blogmarks img{background-color: #535353}.heateor_sss_amp_bookmarks_fr img{background-color: #E8EAD4}.heateor_sss_amp_box_net img{background-color: #1A74B0}.heateor_sss_amp_buddymarks img{background-color: #ffd400}.heateor_sss_amp_care2_news img{background-color: #6EB43F}.heateor_sss_amp_comment img{background-color: #444}.heateor_sss_amp_diary_ru img{background-color: #E8D8C6}.heateor_sss_amp_diaspora img{background-color: #2E3436}.heateor_sss_amp_dihitt img{background-color: #FF6300}.heateor_sss_amp_diigo img{background-color: #4A8BCA}.heateor_sss_amp_douban img{background-color: #497700}.heateor_sss_amp_draugiem img{background-color: #ffad66}.heateor_sss_amp_evernote img{background-color: #8BE056}.heateor_sss_amp_facebook_messenger img{background-color: #0084FF}.heateor_sss_amp_fark img{background-color: #555}.heateor_sss_amp_fintel img{background-color: #087515}.heateor_sss_amp_flipboard img{background-color: #CC0000}.heateor_sss_amp_folkd img{background-color: #0F70B2}.heateor_sss_amp_google_news img{background-color: #4285F4}.heateor_sss_amp_google_classroom img{background-color: #FFC112}.heateor_sss_amp_google_gmail img{background-color: #E5E5E5}.heateor_sss_amp_hacker_news img{background-color: #F60}.heateor_sss_amp_hatena img{background-color: #00A6DB}.heateor_sss_amp_instapaper img{background-color: #EDEDED}.heateor_sss_amp_jamespot img{background-color: #FF9E2C}.heateor_sss_amp_kakao img{background-color: #FCB700}.heateor_sss_amp_kik img{background-color: #2A2A2A}.heateor_sss_amp_kindle_it img{background-color: #2A2A2A}.heateor_sss_amp_known img{background-color: #fff101}.heateor_sss_amp_line img{background-color: #00C300}.heateor_sss_amp_livejournal img{background-color: #EDEDED}.heateor_sss_amp_mail_ru img{background-color: #356FAC}.heateor_sss_amp_mendeley img{background-color: #A70805}.heateor_sss_amp_meneame img{background-color: #FF7D12}.heateor_sss_amp_mixi img{background-color: #EDEDED}.heateor_sss_amp_myspace img{background-color: #2A2A2A}.heateor_sss_amp_netlog img{background-color: #2A2A2A}.heateor_sss_amp_netvouz img{background-color: #c0ff00}.heateor_sss_amp_newsvine img{background-color: #055D00}.heateor_sss_amp_nujij img{background-color: #D40000}.heateor_sss_amp_odnoklassniki img{background-color: #F2720C}.heateor_sss_amp_oknotizie img{background-color: #fdff88}.heateor_sss_amp_outlook_com img{background-color: #0072C6}.heateor_sss_amp_papaly img{background-color: #3AC0F6}.heateor_sss_amp_pinboard img{background-color: #1341DE}.heateor_sss_amp_plurk img{background-color: #CF682F}.heateor_sss_amp_pocket img{background-color: #ee4056}.heateor_sss_amp_polyvore img{background-color: #2A2A2A}.heateor_sss_amp_printfriendly img{background-color: #61D1D5}.heateor_sss_amp_protopage_bookmarks img{background-color: #413FFF}.heateor_sss_amp_pusha img{background-color: #0072B8}.heateor_sss_amp_qzone img{background-color: #2B82D9}.heateor_sss_amp_refind img{background-color: #1492ef}.heateor_sss_amp_rediff_mypage img{background-color: #D20000}.heateor_sss_amp_renren img{background-color: #005EAC}.heateor_sss_amp_segnalo img{background-color: #fdff88}.heateor_sss_amp_sina_weibo img{background-color: #ff0}.heateor_sss_amp_sitejot img{background-color: #ffc800}.heateor_sss_amp_skype img{background-color: #00AFF0}.heateor_sss_amp_sms img{background-color: #6ebe45}.heateor_sss_amp_slashdot img{background-color: #004242}.heateor_sss_amp_stumpedia img{background-color: #EDEDED}.heateor_sss_amp_svejo img{background-color: #fa7aa3}.heateor_sss_amp_symbaloo_feeds img{background-color: #6DA8F7}.heateor_sss_amp_telegram img{background-color: #3DA5f1}.heateor_sss_amp_trello img{background-color: #1189CE}.heateor_sss_amp_tuenti img{background-color: #0075C9}.heateor_sss_amp_twiddla img{background-color: #EDEDED}.heateor_sss_amp_typepad_post img{background-color: #2A2A2A}.heateor_sss_amp_viadeo img{background-color: #2A2A2A}.heateor_sss_amp_viber img{background-color: #8B628F}.heateor_sss_amp_webnews img{background-color: #CC2512}.heateor_sss_amp_wordpress img{background-color: #464646}.heateor_sss_amp_wykop img{background-color: #367DA9}.heateor_sss_amp_yahoo_mail img{background-color: #400090}.heateor_sss_amp_yahoo_messenger img{background-color: #400090}.heateor_sss_amp_yoolink img{background-color: #A2C538}.heateor_sss_amp_youmob img{background-color: #3B599D}.heateor_sss_amp_gentlereader img{background-color: #46aecf}.heateor_sss_amp_threema img{background-color: #2A2A2A}.heateor_sss_amp_bluesky img{background-color:#0085ff}.heateor_sss_amp_threads img{background-color:#000}.heateor_sss_amp_raindrop img{background-color:#0b7ed0}.heateor_sss_amp_micro_blog img{background-color:#ff8800}.heateor_sss_amp amp-img{border-radius:999px;} .amp-logo amp-img{width:190px} .amp-menu input{display:none;}.amp-menu li.menu-item-has-children ul{display:none;}.amp-menu li{position:relative;display:block;}.amp-menu > li a{display:block;} /* Inline styles */ div.acsse3e3c{font-weight:bold;}amp-img.acss334b9{max-width:35px;}div.acss138d7{clear:both;}div.acssf5b84{--relposth-columns:3;--relposth-columns_m:2;--relposth-columns_t:2;}div.acss78e4c{aspect-ratio:1/1;background:transparent url(https://aiofm.net/wp-content/uploads/2025/02/Woman-scammed-out-of-E800k-by-an-AI-deep-fake.webp-150x150.webp) no-repeat scroll 0% 0%;height:150px;max-width:150px;}div.acss6bdea{color:#333333;font-family:Arial;font-size:12px;height:75px;}div.acss386e3{aspect-ratio:1/1;background:transparent url(https://aiofm.net/wp-content/uploads/2024/12/Harvard-Is-Releasing-a-Massive-Free-AI-Training-Dataset-Funded-150x150.jpg) no-repeat scroll 0% 0%;height:150px;max-width:150px;}div.acssd0af2{aspect-ratio:1/1;background:transparent url(https://aiofm.net/wp-content/uploads/2023/10/Machine-learning-reveals-the-contents-of-ancient-scrolls-and-stone.jpg) no-repeat scroll 0% 0%;height:150px;max-width:150px;}div.acssc0b7d{-webkit-box-shadow:none;box-shadow:none;left:-10px;top:100px;max-width:44px;}amp-img.acss8b671{max-width:40px;} .icon-widgets:before {content: "\e1bd";}.icon-search:before {content: "\e8b6";}.icon-shopping-cart:after {content: "\e8cc";}

Selective Forgetting Can Help AI Learn Better

Spread the love

The original version of this story appeared in Quanta Magazine.

A team of computer scientists has created a nimbler, more flexible type of machine learning model. The trick: It must periodically forget what it knows. And while this new approach won’t displace the huge models that undergird the biggest apps, it could reveal more about how these programs understand language.

The new research marks “a significant advance in the field,” said Jea Kwon, an AI engineer at the Institute for Basic Science in South Korea.

The AI language engines in use today are mostly powered by artificial neural networks. Each “neuron” in the network is a mathematical function that receives signals from other such neurons, runs some calculations, and sends signals on through multiple layers of neurons. Initially the flow of information is more or less random, but through training, the information flow between neurons improves as the network adapts to the training data. If an AI researcher wants to create a bilingual model, for example, she would train the model with a big pile of text from both languages, which would adjust the connections between neurons in such a way as to relate the text in one language with equivalent words in the other.

But this training process takes a lot of computing power. If the model doesn’t work very well, or if the user’s needs change later on, it’s hard to adapt it. “Say you have a model that has 100 languages, but imagine that one language you want is not covered,” said Mikel Artetxe, a coauthor of the new research and founder of the AI startup Reka. “You could start over from scratch, but it’s not ideal.”

Artetxe and his colleagues have tried to circumvent these limitations. A few years ago, Artetxe and others trained a neural network in one language, then erased what it knew about the building blocks of words, called tokens. These are stored in the first layer of the neural network, called the embedding layer. They left all the other layers of the model alone. After erasing the tokens of the first language, they retrained the model on the second language, which filled the embedding layer with new tokens from that language.

Even though the model contained mismatched information, the retraining worked: The model could learn and process the new language. The researchers surmised that while the embedding layer stored information specific to the words used in the language, the deeper levels of the network stored more abstract information about the concepts behind human languages, which then helped the model learn the second language.

“We live in the same world. We conceptualize the same things with different words” in different languages, said Yihong Chen, the lead author of the recent paper. “That’s why you have this same high-level reasoning in the model. An apple is something sweet and juicy, instead of just a word.”

Source Link

admin

Recent Posts

Is Your Data Storage Strategy AI-Ready?

The adoption of AI has caused an increased need for proper data governance, and companies…

3 days ago

OpenAI’s new GPT-4.1 AI models focus on coding

OpenAI on Monday launched a new family of models called GPT-4.1. Yes, “4.1” — as…

3 days ago

Evan Brown, Executive Director of EDGE at the Oklahoma Department of Commerce

Evan Brown serves as the Executive Director of EDGE (Economic Development Growth and Expansion) at…

6 days ago

Sex-Fantasy Chatbots Are Leaking a Constant Stream of Explicit Messages

All of the 400 exposed AI systems found by UpGuard have one thing in common:…

6 days ago

AI craze mania with AI action figures and turning pets into people

In the 90s, we collected Pokémon cards, in the 2000s, we all had a weird…

6 days ago

The Medicaid Cut Effect: Can AI Prevent an Incoming Healthcare Crisis?

Medicaid has become a central point of a heated political battle, as Republican lawmakers push…

1 week ago

This website uses cookies.